3D DIGITAL DIMENSIONS 2011 Miami / 26 - 28 October

Conference papers

All rights reserved. No part of this publication may ESOMAR Office: be reproduced, stored in a retrieval system of any Barbara Strozzilaan 384 nature, or transmitted or made available in any Eurocenter 2 form or by any means, electronic, mechanical, 1083 NH Amsterdam photocopying, recording or otherwise, without the The Netherlands prior written permission of ESOMAR. ESOMAR will Tel.: +31-20-664 21 41 pursue copyright infringements. Fax: +31-20-664 29 22 Email: customerservice@esomar.org In spite of careful preparation and editing, this Website: www.esomar.org publication may contain errors and imperfections. ESOMAR Publication Series Volume S250 Authors, editors and ESOMAR do not accept any ISBN 92-831-0254-1 responsibility for the consequences that may arise as a result thereof. The views expressed by the authors in this publication do not necessarily represent the views of ESOMAR.

By the mere offering of any material to ESOMAR in order to be published, the author thereby guarantees: • that the author - in line with the ICC/ESOMAR International Code on Market and Social Research - has obtained permission from clients and/ or third parties to present and publish the information contained in the material offered to ESOMAR; • that the material offered to ESOMAR does not infringe on any right of any third party; and • that the author shall defend ESOMAR and hold ESOMAR harmless from any claim of any third party based upon the publication by ESOMAR of the offered material.

Published by ESOMAR, Amsterdam, The Netherlands Edited by: Deborah S. Fellows

TABLE OF CONTENTS

PROGRAMME COMMITTEE 4

EDITORIAL 5 Reg Baker

PART 1: SETTING THE SCENE 6

Digital trends forecast 2012 7 Dominic Harrison

Eve-olution: Why women rule the web 16 Understanding the digital lives of women around the world Amy Janis, Dan Brilot, Brian Cooper, Sarah Elliott, David Ludica, Tony Marlow

PART 2: ONLINE RESEARCH DIMENSION 27

The researchification of games 28 Adopting a game designer’s approach to market research Peter Harrison

How far is too far? 41 Traditional, flash and gamification interfaces, and implications for the future of market research online survey design Bernie Malinoff, Jon Puleston

Facial imaging 63 The new face of online survey research Alastair Gordon, David McCallum, Matteo Sorci, Tim Llewellyn

PART 3: SOCIAL MEDIA RESEARCH DIMENSION 78

The Rosetta Stone meets Foucault 79 Understanding social media via discourse analysis Ray Poynter

Review of a decade of netnography research 85 Implications for future social media analysis Michael Bartl, Gregor Jawecki, Jan Henric Stönner, Dominic Gastes

Tell me what you want, what you really, really want 94 Creating desired results from social media research Annie Pettit

Leveraging the power of social media 103 The impact of shared news content on consumers Gregg Liebman, Carl Marci

Lessons from the front lines 110 How to engage BRIC consumers in multinational online communities Manila Austin, Katrina Lerman

Breaking the mirror 133 Ideas for an inspiring online dialogue Jon Rodriguez, Isabella Hoi Kee Wong

The earthquake and tsunami in Japan 142 Investigate the possibilities and limits of social media research from researching how the earthquake changed the sense of value of Japanese people Koki Uchiyama, Michio Mutoh, Eriko Ota

PART 4: MOBILE RESEARCH DIMENSION 159

Digital matrimony: marriages that are transforming the face of research 160 Do you, the Internet, take social media and mobile, to converge into one platform? Kristin Luck

Time is of the essence 165 Mobile solutions hold the key to tactical actionable research Adhil Patel

The place for mobile research? 175 Multi-mode studies of major cultural events Simon Atkinson, Sean Conry

PROGRAMME COMMITTEE

Reg Baker (Committee Chair) Market Strategies International, USA

Betty Adamou Research Through Gaming, UK

Jeffrey Henning Affinnova, USA

Dave King Confirmit Mobile Solutions, Canada

Guy Rolfe Kantar Operations, UK

Niels Schillewaert InSites Consulting, Belgium

EDITORIAL 3D DIGITAL DIMENSIONS – (ONLINE + SOCIAL MEDIA + MOBILE) RESEARCH

This is the seventh in a series of ESOMAR-sponsored conferences dedicated to exploring the evolution of market research in an increasingly digital world. Beginning with the Worldwide Panel Research Conference in Budapest in 2005 these conferences have documented an explosion of new research methods like nothing before experienced in our industry. The tools for understanding consumer motivation and behavior have never been more numerous or more powerful, the opportunities for delivering richer insights to clients never more promising.

This conference continues in the tradition of its predecessors as we once again explore the broad landscape of digital market research. Our program is organized around the three main categories of new methods familiar to all of us—online surveys, social media, and mobile. Within each of those categories we have papers that describe the latest innovations and the newest applications. And we have papers that focus on the lessons to be learned from methods that may be relatively new but are nonetheless approaching maturity.

We also will spend some time talking about the context for these new methodologies. While digital technology has been at the heart of much of the innovation in our industry it is only an enabler. The real driver has been the integration of many of these technologies into the daily lives of people worldwide. We should not forget that for all the sometimes magical properties of our digital world it is at its core a means for people to communicate with one another and share information. Understanding how they go about doing that is a key prerequisite of good research.

The essence of market research is delivering insights that enlarge our clients’ understanding of the consumers who make up their markets. Regardless of how our methods evolve or those forces that may seem to be driving that evolution, people, not technology, are at the center of what we do. The challenge to market research in a digital world is to understand how people use technology in their daily lives, how they use it to interact with one another and how we can use it to interact with them. If we have learned nothing else over the last decade we should at least have learned this: doing research with people means doing it on their terms, not ours.

It is not clear to me that we have identified all of the pieces of this puzzle, let alone assembled them into a roadmap to a redefined industry. Like its predecessors, this conference continues that quest. The Programme Committee has worked hard to bring together the best thinking on many of the key challenges our industry faces and the opportunities that lie before us. We hope you enjoy hearing what our presenters have to say, find it stimulating, and will share your own views liberally in the discussions and breakouts.

It’s an exciting time to be working in market research. Let’s enjoy it.

REG BAKER

PROGRAMME COMMITTEE CHAIRPERSON

PART 1: SETTING THE SCENE

DIGITAL TRENDS FORECAST 2012

Dominic Harrison

INTRODUCTION By the beginning of the decade beginning with 2010, the truly global extent of the digital revolution was plain to see: access to mobile networks was available to 90% of the world’s population, including 80% of those living in rural areas; the number of SMS messages sent each year had tripled from levels seen in 2007 to reach some 6.1 trillion; those using the internet passed the 2 billion mark; mobile internet subscriptions had grown more than tenfold globally between 2005- 20101)

As in-home and mobile internet connections proliferate and as the technical wizardry inside digital devices is upgraded with each new incarnation, brands and companies are challenged - on what seems a daily basis - to refresh their understanding of how consumers mobilise digital platforms to, for instance, communicate with each other; to gather global and local information; to augment leisure behaviours; to interact with retail spaces; to play and to shop.

In order to assess and interpret global consumers’ changing relationship with the digital space - and to isolate the key trends central to this developing relationship - the Future Foundation can draw upon its proprietary global research programme, nVision Global, which covers 22 key consumer markets. This proprietary research programme comprises an annual online quantitative survey in each market and an ongoing qualitative research programme that includes online focus groups held in each market as well as a growing network of on-the-ground “trend spotters” tasked with responding to specific assignments, providing branded examples of trends manifesting locally and alerting us to emerging phenomena and behaviours in their home country. nVision Global also draws heavily on an extensive range of complementary external global sources.

In this paper, we focus on four trend themes which we believe will be at the heart of the global consumer’s relationship with the digital world in the years ahead. We will reference our robust research findings throughout, including commercial manifestations of the trends in question. Each trend will conclude with a number of actionable and future-focused insights to identify the potential commercial implications and opportunities each trend represents.

The four themes we focus on in this paper are as follows:

1. Digitally enhanced downtime and the rise of smart boredom 2. Digital etiquette and a future of identity management 3. The integration of the play ethic 4. The emergence of the quantified self

1. Digitally enhanced downtime and the rise of smart boredom Our first trend theme explores the notion that digital advancements are contributing to a radically altered consumer approach to the concept of downtime. Increasingly so, digitally-equipped consumers are able to use moments of downtime constructively by undertaking low-level activities that continue to add value to their lives in one form or another.

As a result, moments once considered to be little more than “dead” time - for instance, the time spent commuting to work, standing in line at the grocery store or passively watching TV at home - are rarely completely vacant at all but can be occupied by ultimately rewarding and often useful activities such as browsing the internet for information, using a smartphone or tablet app to check a bank balance or playing a game on one of any number of sophisticated portable devices available today.

This is a trend that has been massively facilitated by digital innovations that enable us to live in a constantly connected world, continuously providing us with ever more convenient solutions for so many daily tasks. The proliferation of mobile

devices, and the mobile internet in particular, is opening up the possibility for millions to enrich any spare moment with purposeful activity. (See figure 1.)

FIGURE 1, FREQUENCY OF MOBILE INTERNET USAGE

Source: nVision Research. Base: 800-4,000 online respondents per country aged 16-64 (*China 16-54), 2010

A significant proportion of internet users are already breaking free of the wired web and exploring all that mobile internet access has to offer. Outside of Europe, we see relatively high usage of the mobile web in China, Japan and India (although we must caveat this with a methodological note: given living conditions in China and India, online samples tend to capture urban, affluent consumers rather than a representative sample of the entire population).

As familiarity with and availability of the mobile internet grows, we can expect engagement with the medium to mature also. In the UK, nVision research suggests that over time, mobile internet users have becoming increasingly likely to carry out more advanced activities on their handsets. Indeed, between Summer 2010 and Spring 2011, mobile-based activities such as “downloading applications”, “checking your favourite social networking site” and “playing games on a social networking site” all predictably grew in popularity. (See figure 2.)

FIGURE 2, PREFERRED THINGS TO DO WHILE WAITING FOR A BUS/ TRAIN

Source: nVision Research. Base: 2,000 - 5000 online respondents aged 16+, GB

As mobile engagement matures so too can we expect mobile activities to feature in those everyday contexts in which we find ourselves at a loose end, looking for worthy distraction in moments of planned or enforced downtime. For many global consumers, such moments are ripe for improvement: indeed, at least four in ten people in all the countries we survey -

with the exception of Germany, Sweden, the Netherlands and Denmark - that admit that one of the things they dread the most is to be bored or to have nothing to do. Mobile platforms’ growing role in augmenting moments of underutilised downtime is revealed in recent research carried out in the UK. Respondents were asked what they would typically prefer to do if they were faced with a 20 minute wait for a bus or a train. Whilst the most popular response in 2011 is “nothing / just relax” (as it was in 2009 when this question was last put to UK respondents), we find that interest in phone-based activities has grown significantly (and particularly so amongst the youngest respondents), including “browse the internet on your phone”, “check your email on your phone” and “read the news online on your phone”.

Conversely, the proportion claiming they would prefer to “call a friend/family member on your mobile phone” has fallen, as has the proportion preferring to “read the newspaper”, suggesting a growing preference for more low-involvement activities - but no less practical or indeed informative - which can so conveniently be delivered through the mobile handset. (See figure 3.)

FIGURE 3, PROPORTION INTERESTED IN USING A DEVICE/APP THAT “ALLOWED ME TO WATCH TELEVISION WHILE OUT AND ABOUT”

Source: nVision. Base: 800-4,000 online respondents per country (**GB sample is 5000) aged 16-64 (*China 16-54), 2010

There is also significant latent demand for more advanced mobile activities not currently available in all global markets. For example, it is the majority of relatively affluent, urbanised respondents in BRIC (Brazil, Russia, India and China) markets who would welcome a device or app that “allowed me to watch television while out and about”. It is over a quarter of consumers in Great Britain, Japan and the USA who would also welcome such innovation in their lives.

We foresee a future in which spare leisure moments, no matter how relaxing or restful, can be intelligently converted into some form of achievement or purposeful activity. We characterise this shift in a trend we call “Smart Boredom”; whilst underpinned by technological innovation, this is, we suggest, a trend symbolic of a wider evolution in consumer attitudes towards downtime and leisure time in general. The desire to efficiently manage the time available to us and to enrich spare moments with smart and self-improving activity will become a key characteristic of the consumer mindset.

Scanning the globe, we notice a number of brands and companies - and from a variety of sectors - creating offers designed to help consumers achieve small efficiencies during moments of the day one would not traditionally associate with personal productivity.

South Korea’s Home Plus, a retail chain jointly owned by Tesco and Samsung (with Tesco owning the majority stake) launched a campaign in 2010 enabling people to shop while on their daily commute. Home Plus installed backlit boards on subway station platforms featuring images of popular groceries. Using a smartphone, consumers scan QR codes next to individual products to add them to a shopping basket. Once paid for, orders are processed and deliveries can be made within a couple of hours. More, if a shopper’s train arrives before they have scanned everything they need, they can continue shopping on the mobile website. Home Plus claims the campaign helped online sales increase dramatically2)

In April 2011, Becks launched an Art Crawl campaign to celebrate what it called “25 years of iconic art labels” featured on its bottles. During a two-week period, eight pubs and bars in East London exhibited artworks by a range of leading figures, including Tracey Emin. In addition, a number of digital screens erected throughout the local area allowed passers-by to

plug in headphones and listen to interviews with a range of artists. Bus-shelters, meanwhile, featured digital maps to guide passers-by to the venues participating in the scheme. This is an interesting example of a brand engaging consumers during fleeting moments of otherwise underutilised downtime, in this case while waiting for public transport.

Actionable insights for the future: Enhanced downtime and the rise of smart boredom  Product and service innovations that aid consumers to complete chores with minimal effort, reward them with a sense of accomplishment or liberate time for other activities chime perfectly with the Smart Boredom sentiment.

 Mobile technology and internet uptake will be key facilitators of the trend. As more and more consumers gain access to smartphones and broadband internet, their ability to engage in non-taxing activities and fill their downtime with easy-to-complete tasks will naturally expand. As a corollary, so too will they expect offers to respond to this newfound expectation.

 Gaming has become the largest entertainment industry in the world and presents consumers with an alluring outlet for Smart Boredom entertainment by offering them cognitive stimulus as well as mere distraction. Games are becoming more accessible across a range of platforms; location-sensitive games allow consumers to play while out- and-about and are increasingly integral to the social network experience for many people. We can engage with games in both fun and serious ways practically anywhere, anytime. We explore gaming in more detail later in this paper.

 Whilst we acknowledge that consumers will always crave entertainment that is frivolous and distracting (at least at some points in their week or day) it is likely that consumer interest in leisure (even passive forms of leisure such as watching TV), will increasingly be geared towards offers that can be seen to be adding significant value to their lives. Purposeful leisure will soon be considered a significant premium.

2. Digital etiquette and a future of footprint management The emergence of the trend towards Smart Boredom implies a future in which engagement with digital platforms and services can only grow - and at all times of day, as digitally-equipped consumers realise the practical and enriching benefits of multiple digitally-enhanced moments. In this second theme, we explore some of the broader ramifications of a digitally-exposed existence, and argue that a new Digital Etiquette is fast emerging within the consumer mindset as many come to realise the value of more closely managing rapidly expanding digital footprints.

Communication culture has been seriously re-formatted by the fixed and mobile internet and it is no great exaggeration to say that day-to-day human relations are being dramatically re-ordered. So much formality has been lost; so much language has been compressed to its vowel-free essentials; so many critiques (about absolutely anything) can be logged; so many images can be tagged; so many indiscretions laid bare; so many things can be said/sent/posted/blogged - and not easily retrieved. Today, many millions are adding to the global digital stockpile on a regular basis; indeed, it is now a majority of social networking 16-24 year-olds in all of the markets we survey that claim to update their social networking profiles or send messages via the medium at least once a week. (See figure 4.)

FIGURE 4, UPDATING PROFILES OR SENDING A MESSAGE ON A SOCIAL NETWORKING SITE AT LEAST ONCE A WEEK

Source: nVision Research. Base: 300-3,000 social network site users per country aged 16-64 (*China 16-54), 2010

As the digital revolution progresses, we notice that nobody has, along the way, promulgated a new good practice handbook; the internet has no government and no church and no department of moral philosophy. But, the consumer- citizen is, we contend, beginning to recognise just how much self-awareness is required in cyberspace (where everyone really can hear you scream) and indeed, just how much value there is in pursuing a more disciplined and considered approach to personal identity management.

Awareness of the non-eradicative permanence of what we say online is encouraging a culture of digital housekeeping. In order to manage the infinite lifespan of certain communications such as ill-advised emails, misjudged comments, perhaps unsavoury body-part photos and more broadly every tracked detail of our past online activities, the need to sweep our digital footprint is likely to become an increasingly widespread lifestyle priority.

Pruning our personal profiles, for instance, is already a habit for those keen to keep presenting themselves in the most attractive or positive light. It is some 40% of 16-24s in the UK, according to nVision - the most active social networking group age - who agree that “I have in the past posted messages/updates on a social networking site which I later wish I had not posted”.

And we note here too the numerous services available that are designed to present the Netizen in his or her best visual light. Consider the successful Chinese site Meitu Xiu Xiu, for instance, which offers browsers the opportunity to download proprietary software (free of charge) that can be used to edit and enhance images posted online - including a “Beauty” feature which appears to help users remove blemishes from photos or even resize facial features - all in the name of digital reputational (and personal image) management. The site claimed (in Spring 2011) to have over 300m active users.

Consider too the slightly tongue-in-cheek Mail Goggles, a Google Labs add-on for Gmail users. It essentially acts as a security measure for those people sending late-night emails when perhaps over-worked or over-refreshed. When it reaches a certain time in the evening the software asks people to solve a handful of math problems before they can hit the Send button.

A concept called Vanish, designed by scientists at the University of Washington, helps with the task of deleting personal digital information when its author no longer wishes to see it online. The service effectively puts an expiry date on personal data released online; after a certain period, data becomes unreadable (un-lockable with encryption software owned only by the data owner).

Footprint management is also a theme alive inside the regulatory agenda of the European Union (EU). Viviane Reding, the EU's Luxembourgian Justice Commissioner, has long called for an overhaul of the EU’s data protection legislation, with a core aim to secure people’s “right to be forgotten” on the internet. The planned legislation would, if enacted, secure the right of those leaving social networking sites to take personal data along with them, to “digitally disappear”.

Actionable insights for the future: Digital etiquette and a future of footprint management  The more digitally sophisticated we become, the greater our awareness of all the pitfalls a rich digital life can entail. We can expect that as our experience of online communications grows across the world, the general human approach to social networking communications will become more considered. Some will define social limits as to what level of digital interaction is appropriate and we think that more universal rules for the digital space (and how digital technologies are used in the real-world) will gradually emerge.

 Ever more advanced tools and services which allow us to regulate our own cyber-presence, image and reputation will surely emerge to help Netizens lead neater, more professionalised lives online. Online identity management will become a common aspiration.

 Amidst all of this, do we see any major disengagement from the digital revolution and its entire works? No. A majority of consumers we survey in Brazil, Argentina, South Korea and India - as well as in several European nations - agree they “never turn off their mobile phones nowadays”; a majority of internet users tell us they engage with social networking in one form or another on a daily basis. We do not sense much digital fatigue amongst global consumers; rather, we anticipate a growing desire to take a more considered, regulated approach to some of our new digital behaviours.

3. The integration of the play ethic The third theme of this paper focuses on the opportunities emerging from the newfound consumer respect for ludic interludes in everyday life; both in the form of electronic gaming (in fixed and mobile locales) and through the integration of playful features into products, services and, increasingly so, marketing campaigns.

Humans have long made room for ludic experience and the idea of the Play Ethic is not a new one (Kane 2004), but like so many of the trends we observe, an underlying tendency has been brought to the fore by a combination of technological advancements and a greater awareness that “play” in many forms may contribute to wellbeing (as an antidote to life’s stresses and underlying pressures) and improved mental capacity.

Author Clay Shirky’s 2010 book Cognitive Surplus (Shirky 2010) argues that new technology is creating additional capacity to interact, communicate and create. Whilst we cannot verify such claims through our research, it is certainly the case that a growing number of people now have access to, and are using, gaming technologies as an integral part of their daily lives - very often as part of a new portfolio of downtime behaviours enjoyed in line with the Smart Boredom trend.

Usage of play technologies has become widespread, and significant proportions of global consumers claim to play video games at least once a week. Important to note, our research challenges the traditional perception that electronic gaming is first and foremost the preserve of young men. Rather, our research suggests that the play ethic has become a vibrant characteristic of many demographic groups.

Whilst consumers in the 16-24 age cohort are most likely to play video games at least once a week, sizeable proportions of older generations and women also claim to play. In part, we suggest, a result of efforts by the electronic games industry to attract so-called “casual gamers” with family-friendly titles and new peripherals, gaming has become a more mainstream and, we suggest, a more highly regarded and acceptable activity. (See figure 5.)

FIGURE 5, PROPORTION PLAYING COMPUTER/VIDEO GAMES AT LEAST ONCE A WEEK IN SELECTED EUROPEAN COUNTRIES

Source: nVision Research. Base: 800-4,000 online respondents per country (*GB sample is 5000) aged 16-64, 2010

Gaming as we have traditionally known it is changing.

The internet is fast emerging as a play destination as games hardware and software increasingly offer the option of online play against others. One area that has grown dramatically is “social gaming”, casual online games that operate entirely within social networking communities and are played amongst friends, placing games at the heart of online social interaction. Such offers have rapidly captured the imagination of global social networkers; close to eight in ten Chinese social networkers, our research confirms, claim to have played a game on a social networking site.

More, as mobile devices become ever-more sophisticated and user-friendly, gaming-on-the-go is evolving into a popular leisure activity, enabling consumers to fill moments of otherwise unproductive downtime. Over a fifth of consumers in Great Britain, Spain, Japan and China claim to play games on their mobile phones at least once a week.

Innovative brands are already finding ways to engage people more deeply by infiltrating existing gaming spaces, which gamers believe makes these more realistic (indeed, branded involvement in games is welcomed by a robust minority in developed markets on the condition that in-game worlds are made to feel more authentic as a result) and by inviting them to play in ways that are relevant to the brand.

There is a wider trend towards gaming attributes finding their way into products, services and marketing campaigns. Manifestations of this trend have been identified by our trend spotters from around the globe and span a number of sectors.

In an interesting confluence of the Smart Boredom and Play trends, November 2010 saw Yahoo equip 20 bus shelters in San Francisco with 72-inch interactive touch-screens that commuters and passers-by could use to play "Bus Stop Derby", a collection of games designed to entertain waiting travelers and connect them with fellow commuters in the San Francisco area. Players rack up scores and compare their progress - in real-time - with those in different parts of the city. An added motivation to play well, the neighbourhood with the highest score at the end of the campaign (January 2011) was rewarded with a special “block party”, with the promise of an appearance by the US band OK Go.

In another example, which sees play being used as a mechanism to incentivize environmentally-friendly behaviour, Toyota Sweden released A Glass of Water, an iPhone app, in late 2010 that encourages people to drive in a more efficient way - for the benefit of both the planet and the driver’s wallet. The app simulates a glass of water perched on the dashboard and drivers are challenged to drive in a way that minimises “spillage” from the glass (i.e. gentle acceleration, smooth breaking, slower speeds, etc.). The app also calculates fuel consumption and driving distance, the results of which are uploaded to Toyota’s mini-site (available in English, Swedish and Norwegian iterations) and compared against those of other app users, thereby introducing a competitive, playful element to the otherwise serious business of improving fuel efficiency.

Actionable insights for the future: The integration of the play ethic  By including game-like attributes into commercial offers and marketing vehicles, brands can tap into so many consumer desires: for social competition and interaction, for personal challenge, motivation and status-enhancing rewards and for good old-fashioned escapist and informal fun.

 As the decade develops we expect electronic gaming to appeal to an ever broader demographic base. Such is the pace of innovation within the gaming industry - e.g. the introduction of family- and friend-oriented console peripherals and a rapidly growing online social gaming sector - that consumers from across the age and gender divides will find much to enjoy in solitary or shared electronic gaming.

 We anticipate a future in which gaming and other entertainment activities increasingly merge. Current generation games consoles - often connected to the main screen in the home - for instance, fulfil a number of different functions, from music and Blu-Ray player to TV and video/game-on-demand conduit, from web browser and photo album to VoIP platform and social networking device.

 A new relationship between brands and consumers will emerge; one in which brands become play providers by embedding play mechanics into offers to entertain users, to nudge them to fulfil personal ambitions and to deliver complex information in a more appealing and digestible format.

 We anticipate that playful mobile games which invite the player to interact with the physical environment surrounding her (e.g. through internet-driven location-based software and augmented reality applications) could become a widespread means to engage consumers with both physical retail locations and outdoor marketing campaigns.

4. The emergence of the quantified self The final theme of this paper explores a relatively nascent digital trend that sees consumers increasingly able to use sophisticated mobile (and fixed) technologies and applications to intelligently amass and analyse the data that surrounds their everyday lives.

We face a future in which consumers are able to scrutinise - with relatively minimal effort and at affordable cost - the impact and efficiency of his or her daily endeavours. With the precise delivery of real-time results, health can be monitored, eco-behaviour tracked, household expenditure managed.

Whilst this is a trend made possible by technological developments - and the availability of the various apps and devices that are able to monitor, for example, personal health metrics or household energy consumption - it connects with a number of powerful underlying consumer aspirations; for instance, the aspiration to lead a fit and healthy existence (voiced

by large majorities in each market we survey) or the aspiration, strengthened by the global economic downturn, to more closely monitor household budgets. The proportion of consumers regularly using personal monitoring apps is small; less than 10% of respondents in all markets surveyed claim, for instance, to use “Personal health tracking apps (e.g. calorie counters, running route markers)” on a monthly basis. However, latent consumer demand for such services is significant: over four in ten respondents in Ireland, Great Britain, Poland, India, Brazil, China, South Korea and Argentina agree that they would be interested in a device or app “that allowed you to track the number of calories/nutritional value of the food you eat”, with interest highest, perhaps unsurprisingly, amongst female and younger groups. (See figure 6.)

FIGURE 6, INTEREST IN NUTRITIONAL MONITORING APPS

Source: nVision Research. Base: 800-4,000 online respondents per country (**GB is 5000), aged 16-64 (*China is 16-54), 2010

Particularly as smartphone penetration soars throughout the 2010s, the amount of instant information available at our fingertips is set to become colossal - highly personal and purposefully packaged too. By mid-2011 a large number of services existed, inviting consumers to monitor various aspects of everyday life.

In an example from the UK that speaks to consumers’ health concerns, the Mydrinkaware app (launched in March 2011 by alcohol awareness charity Drinkaware) is a free and confidential tool which allows users to track their weekly intake of alcohol and provides, its developers claim, “instant feedback on how many units you are consuming, how many calories too, and how much you're spending - helping you stay healthy and in control”. Those who download the app are invited to keep a weekly diary which can then be compared against recommended allowances - with the charity hoping it will make individuals aware of the dangers (and financial costs) that excessive consumption can bring.

Push Snowboarding, initially a collaboration between telecommunications company Nokia and sportswear manufacturer Burton, is an ongoing project that provides snowboarders with advanced equipment that can monitor physical and mental performance during a snowboard run. The project revolves around five measurements, taken by the rider’s smartphone and by sensors located around their outfit. For instance, the “rush” metric - which indicates how well the rider is holding his or her nerve - is measured by galvanic skin response sensors; the rider’s speed is calculated using a Nokia N8’s in- built GPS function. Ultimately, the collected information presents the rider with an accurate log of personal performance which can be interrogated to unearth valuable recommendations on how to improve, for example, posture or mental preparedness. Is this the future of high-performance sportswear?

Actionable insights for the future: The emergence of the quantified self  The desire to bring all important life-facts under personal control will only grow stronger. Driving this will be: a) the ever improving scope and sophistication of the real-time information available to us, facilitated by technological development; b) the underlying consumer desire to find the best deal, the most energy-efficient option, the healthiest choice; and c) an increasing rejection of the notion that ignorance is bliss.

 Especially in developed markets where public finances are tight and the post-recession psyche is such that millions search for every possible household efficiency, the pressure to run responsible lifestyles will intensify. Indeed, hyper- intelligent lifestyle management will, we anticipate, become a highly valued skill - one which people will parade via their social networks as cultural capital.

 By providing attainable targets and implementable advice, self-tracking devices could engage customers and even stimulate brand loyalty. Incorporating a competitive element - by, say, allowing consumers to compare the performance of their cars or heating systems or low-energy appliances to those of their peers - could become a key form of incentivisation.

 There will be an increased expectation that companies will strive to continuously improve their offers - the car must be ever more fuel-efficient, the ready-meal must contain ever reduced levels of trans-fats. Quite simply, self- monitoring devices will make customers so much better at exposing the underperforming offer to themselves. The revolutionary impact of The Quantified Self on marcoms everywhere cannot be questioned.

 But few global consumers, we posit, will want to fill their houses with an array of tracking apps. Synchronised, passive devices - which can process various datastreams and offer additional benefits by, say, transmitting health- readings to our doctor or automatically switching off wasteful appliances - will fit best with the groove of this trend.

CONCLUSION In this paper we have introduced a selection of the key themes driving consumer engagement with the digital space in the 2010s decade. A complex consumer relationship with the digital realm is set to develop; a relationship as purposeful as it is playful, as considered as it is quantifiable.

NOTE This paper is primarily based on global research carried out by the Future Foundation in late 2010. Findings from the most recent 2011 wave of research, which was at the fieldwork stage of collection at the time of writing this paper, will be presented by the author at 3D Digital Dimensions 2011. As such, the content of this paper may not fully reflect that which is presented at the conference, yet is designed as a guide to the salient themes shaping global consumers’ digital behaviours in the early 10s decade and beyond.

FOOTNOTES 1. UN ITU World Telecommunication/ICT Indicators database 2010 2. Between November 2010 - January 2011, Home Plus claimed the number of registered users rose by 76% and online sales grew by 130%.

THE AUTHOR Dominic Harrison is the Head of Global Trends, Future Foundation, United Kingdom.

EVE-OLUTION: WHY WOMEN RULE THE WEB UNDERSTANDING THE DIGITAL LIVES OF WOMEN AROUND THE WORLD Amy Janis • Dan Brilot • Brian Cooper • Sarah Elliott • David ludica • Tony Marlow

THINK PINK: WOMEN AND THE WEB The Internet has grown by leaps and bounds, opening up opportunities for people to connect across a variety of activities, from dating to advertising to blogging. The Internet evolution has proven to be something of an Eve-olution, with women playing a pivotal role, satisfying their need for connection.

Once viewed as the rarified domain of male teenage “gear-heads”, the Internet today boasts some 2.1 billion users worldwide1) and almost half of the online population is now female.

From inception, gender differences in web activities were apparent. Men tended to download software programs and viewed the web as a form of amusement during leisure hours (anyone remember Zork? Doom? Space Invaders?). Women turned to the web as an efficiency booster, maintaining family schedules and searching magazine, health and retail sites for relevant content to cope with daily life. This paper looks at the current state of gender-based differences related to online activity, investigating questions like:

 What are the online behavioral differences between men and women?  Why do women use different online channels and what specific needs do they fulfill?  How can advertisers leverage this information to be more effective with their digital campaigns?

Women in general, and mums in particular, represent high-value targets for advertisers. The studies underpinning this analysis provide tone and content strategies for digital ad campaigns with the potential to deliver maximum marketing resonance, rendering brands and advertising more relevant, engaging and effective.

GIRL POWER Although the Web is still in its relative infancy for emerging countries, significant demographic differences have surfaced in digital media utilization. Among the most dramatic differences are those associated with gender. Drawing on a number of studies and data sources, the authors have determined that content appealing to men may hold little interest for women, and digital strategies that may succeed in one country cannot be ported automatically to another.

Women have expanded their sphere of influence to the digital mainstream, accounting for almost half (46%) of global Internet traffic.2) Women comprised more than half of the 215 million U.S. Internet users as of July 2011 and more than half of all North American users.3)

Even in areas where women represent a minority of the online population, they’re closing-in fast on the top spot. Presently, women make up 50% of Latin American, 49.5% of European, and 43% of Asia-Pacific Internet users. In terms of sheer economic clout, those percentages convert into 45.3 million Latin women, 162.8 million European women and 206.4 million Asian women consumers.4)

TICK TOCK On a worldwide basis, an inverse relationship exists between the number of women on the web and the amount of time spent online. While there are fewer women than men active online globally, women spend more time per person on web- based activities (24.8 hours per month per woman) than men (22.9 hours per man).5) Some regional patterns differ, with men recording more online hours than women in Canada, the United States and Latin America (see figure 1).

FIGURE 1, AVERAGE HOURS SPENT ONLINE PER MONTH BY REGION AND GENDER

Source: comScore Media Metrix, February 2011

To visualize this trend on an indexed regional basis, it becomes apparent at a glance that women devote more time to online pursuits, most notably in the Asia Pacific region.

DIFFERENT STROKES In the past, marketers assumed that more men than women were technology early adopters and targeted this elite corps of gadget gurus to achieve rapid uptake of new technologies. While that may have been true once upon a time, the premise no longer holds.

The Digital Life study examined different types of online activity, comparing usage patterns by gender, and determined that women have equaled or surpassed their male counterparts in eight of 11 distinct areas. They spend 36 more minutes per week on social media than men, 18 more minutes shopping and six minutes more on email, organizing and personal interests. Younger women ages 16-20 in particular spend a significant amount of time on social media, logging 6 hours 48 minutes per week versus 5 hours for the average woman or 4 hours 24 minutes for men. (See figure 2.)

FIGURE 2, HOURS PER WEEK OF DIGITAL ACTIVITY BY GENDER

Source: TNS Digital Life Study, 2010

Internet interest patterns for men encompass online gaming (48 more minutes per week than women), catching up on news (42 more minutes per week than women) and multi-media viewing (six more minutes per week than women). Both genders spend an equal amount of time in knowledge-based online activity, administrative tasks and general browsing.

LEADING CATEGORIES, SITES Where do women spend their online time? What destinations draw them in and keep them engaged? Are women and men really that different? Perhaps -- when it comes to needs states and motivation; less so when it comes to the most popular web categories.

The list of Top 10 categories for adult men and women around the world is identical. The only difference is the top two positions change with gender: women spend the most time on web portals, while services sites such as message boards, chat groups, coupon/incentive sites and email occupy the top spot on the list for men. (See table 1.) TABLE 1, TOP 10 WEB CATEGORIES FOR ADULT FEMALES 18+ YEARS

1. Portals (#2 men) 2. Services (#1 men) 3. Corporate Presence 4. Search / Navigation 5. Entertainment 6. Conversational Media 7. Promotional Servers 8. Social Networking 9. Directories / Resources 10. News / Information Source: comScore Media Metrix, June 2011 In terms of specific web sites, the six most popular web sites for American men and women were the same: Yahoo!, Google, Microsoft, Facebook, AOL and Amazon. Wikimedia and Turner Digital slotted onto both lists, but in different rankings among the Top 10 sites. The men-only sites listed were CBS Interactive and eBay, while the women-only sites completing the Top 10 list were Ask Network and Glam Media.6)

THE GLOBAL SHOUT-OUT Relationships. Connections. Affiliations. Associations. Friendships. Alliances. Unlike men, women around the world share a common psychology that emphasizes frequent and emotionally-resonant communication across all available outlets.7) This stands as a foundation pillar supporting the degree and intensity of the female involvement with online media. In the United States alone there are 2.1 million more women than men who communicate online regularly.8) (See figure 3.)

FIGURE 3, U.S. ONLINE COMMUNICATION PATTERNS BY GENDER

Source: comScore Media Metrix, July 2011

Looking across all online activity on a worldwide basis, women spend half their online time communicating via instant messaging, email, discussion/chat rooms, ecards, social networking, blogs and community sites, while men spend just 44% of online time in those pursuits.

While the general gender communication trend holds constant around the world (women communicating online more than men), regional differences emerged. Latin American women devoted the most time of any region to digital discussions, more than two-thirds of their online time, followed by European women at 54%, U.S. women at 45% and Asia Pacific women at 38%. (See figure 4)

FIGURE 4, REGIONAL VIEW OF TIME SPENT COMMUNICATING X GENDER

Source: comScore Media Metrix, February 2011

Women tend to engage in certain types of online “connecting” activities more than men. Taking pictures (figure 5) and sending mobile messages to friends (figure 6) are quite common experiences for connecting consumers. Worldwide, 27% of females take and share pictures with their friends versus 21% of men. Women are also much more likely to communicate with their friends via text messaging on their mobile phones across both emerging and developed markets. Regardless of the type of connection that people make, females tend to drive those personal connections much more than men.

FIGURE 5, PERCENT OF MEN/WOMEN SHARING PICTURES

Source: TNS Digital Life Study, 2010

FIGURE 6, PERCENT OF MEN/WOMEN SENDING TEXT MESSAGES

Source: TNS Digital Life Study, 2010

RETAIL THERAPY One of the unique ways in which women engage the Internet is their use of retail therapy. Whether perusing brick and mortar outlets or online shopping options, nothing lifts the spirits more than a little “retail therapy”. Popular culture portrays shopping as a purely female pastime, but the data prove otherwise. While it is true more women than men shop

online in the United States (86.8 million vs. 84.7 million respectively), male bargain hunters spend 151 more minutes per year checking out shopping sites for a total of 2,432 annual Internet shopping minutes per man on average.9)

On a global basis, women clock significantly more online shopping minutes than men, particularly women ages 25-34 who spend 13.8 more minutes per month than their male counterparts, followed by women ages 45-56 who log nine more minutes per month. (See figure 7.)

FIGURE 7, GLOBAL ONLINE RETAIL TIME/ MINUTES PER PERSON BY GENDER

Source: comScore Media Metrix, June 2011

Men and women exhibit very similar online retail patterns, sharing nine of the Top 10 shopping sites, albeit with different rank orders. The leading male sites skew towards entertainment and electronics, the more popular female sites tend towards more general merchandise or family-oriented options. (See table 2.)

TABLE 2, TOP 10 INTERNET RETAIL SHOPPING SITES FOR U.S. BY GENDER (JULY 2011)

Site Ranking Women Ranking Men Amazon Sites 1 1 Wal-Mart 2 3 Apple.com Worldwide Sites 3 2 Target Corporation 4 8 NETFLIX.COM 5 4 Yahoo! Shopping 6 5 Hewlett Packard 7 10 Shopzilla.com Sites 8 - Google Shopping 9 6 Nextag.com Sites 10 9 Best Buy - 7 Source: comScore Media Metrix, July 2011

Breaking down Internet shopping behavior by category, women out-shop men in 14 of the 20 largest categories: comparison shopping, apparel, books, department stores, home furnishings, flowers/gifts/greetings, fragrances/cosmetics, mail, health care, tickets, food, consumer goods, jewelry/luxury goods and toys. Consistent with gender stereotypes, men gravitate toward computer hardware, consumer electronics, computer software, sports/outdoor interests, movies and music purchases.10) (See figure 8)

FIGURE 8, GLOBAL RETAIL SUBCATEGORY/REACH BY GENDER

Source: comScore Media Metrix, June 2011

MEET DR. MOM Another unique function for digital women is the role of the health care provider. Call her “Dr. Mom”, “Caregiver-in-Chief” or “Chief Wellness Officer”, but whatever honorific you prefer, there’s no denying that women serve as the family health gatekeeper, significantly more immersed in health and wellness content online than men at every age. The time women spend online per year investigating health and wellness information begins to ramp in their twenties and peaks during middle age at more than 22 minutes per person per month, reflecting their enlarged circle of responsibility. (See figure 9)

FIGURE 9, GLOBAL HEALTH BY AGE, GENDER: % REACH, TIME SPENT

Source: comScore Media Metrix, June 2011

These midi-life mums keep tabs on everyone’s physical state - their kids, their spouse, their elderly parents, even the grandkids. More than 86.8 million U.S. women turn to the web for health advice, spending 2,281 minutes per person annually on average.11)

Globally, women’s interest in health issues increases over time. The gender gap opens up at 3% among the 15-24 age cohort, widening to 10% by age 55, with 42% of women exploring topics ranging from diet and exercise to disease states

and treatments versus just 32% of men. Interestingly, men’s interest in health issues peaks at ages 35-44 while women’s concern grows steadily over time, accounting for 22.8 online minutes per month by the time she reaches age 55. (See figure 9.)

EMOTIONAL CONNECTIONS The type of personal connections women have with the Internet are not universal. When developing campaigns it is important to note that there are strong regional differences worldwide. Not only are there strong behavioral differences between men and women, but the emotional connections women have with the Internet vary quite dramatically by emerging versus developed markets. Women in developing markets are more than twice as likely as peers in developed nations to say they express their feelings more freely in the online world where political, cultural and religious strictures do not apply. (See figure 10.)

FIGURE 10, THE INTERNET AS A MEDIUM FOR PERSONAL EXPRESSION

Source: TNS Digital Life Study, 2010

There are many facets to this trend. Relative newcomers to the online world tend to access the web via mobile devices which lend themselves to shorter, timelier connections with others online. Emerging market users relate to the Internet in a more aspirational than functional way, perceiving it as a vehicle for achieving goals and pivotal to their life. Given women’s high level of receptivity to the Internet in emerging markets, this presents significant opportunities for advertisers to take advantage of.

ADVERTISING RECEPTIVITY Extending these personal connections into advertising campaigns is extremely important. Breaking through online clutter with a personally-resonant and contextually-relevant advertisement is critical to maximizing the return on investment (ROI) of an advertising campaign everywhere in the world. Across a variety of media and devices, women are much more likely to be receptive to advertising campaigns than men. Coupled with female online activity levels, this bodes well for advertisers looking to obtain a higher ROI with their digital campaigns.

Females report a higher likelihood of advertising receptivity in pre-purchase/browsing and shopping activities. Women in emerging markets are twice as likely as developed market females to exhibit a higher propensity to advertising receptivity during the online pre-purchase and purchase phases. (See figures 11 and 12.)

FIGURE 11, ONLINE PRE-PURCHASE AND BROWSING AD RECEPTIVITY

Source: TNS Digital Life Study, 2010

FIGURE 12, ONLINE PURCHASE AD RECEPTIVITY

Source: TNS Digital Life Study, 2010

EVERYONE’S CONNECTED Extending these personal connections into advertising campaigns is extremely important. Breaking through online clutter with a personally-resonant and contextually-relevant advertisement is critical to maximizing the return on investment (ROI) of an advertising campaign everywhere in the world. Across a variety of media and devices, women are much more likely to be receptive to advertising campaigns than men. Coupled with female online activity levels, this bodes well for advertisers looking to obtain a higher ROI with their digital campaigns.

Even the visionaries responsible for creating the web could never have conceived of the Internet’s ultimate global impact. The web functions as electronic connective tissue binding humanity together into a living organism that transcends the artificial boundaries of geography, time, gender, culture or religion that once separated humankind.

For women, connectivity assumes even greater importance. The female need to communicate is paramount, accounting for the largest share of her online time, regardless of lifestage or demographic variable. Communication activities such as social networking, email, blogging and instant messaging comprise 52% of online time for women around the globe.

Marketers looking to engage with women via social media can follow a simple formula: ME = f (N + C + R) where online marketing effectiveness (ME) with women is a function of meeting needs (N), selecting appropriate channels (C) and matching the message to a receptive (R) outlet.

NEEDS: WHY WOMEN COMMUNICATE ONLINE The Yahoo! Connectonomics: U.S. research studied 3,000 American women online and discovered 16 specific need states that explain the majority of their online behavior: care of self (83%), bargain hunting (74%), improve myself (72%), be enabled (71%), affectionate closeness (64%), mutual sharing (61%), release and escape (58%), broadening horizons (50%), validation (41%), repair and healing (38%), in control (37%), be in the know (29%), being the best (18%), up the ante (10%), social currency (8%) and pushing the limits (5%).

After analyzing these findings by standard demographic variables such as age, employment status and presence of children, it became evident that the need state drivers behind female online communication patterns surmount age groups, lifestages and geography.

Simply put, the reasons women communicate online are more similar than different. To illustrate, regional Connectonomics studies determined that the need states of women in Brazil, Mexico, Saudi Arabia, Egypt and the United Arab Emirates all converge around three areas: self-improvement/personal growth, mutual sharing/ bonding and being enabled/informed.

CHANNELS: COMPLEX LIVES, SIMPLE NEEDS Women lead incredibly complex lives with increasingly conflicting responsibilities. (See figure 12.) Mother, spouse, daughter, sister, employee, Momtrepreneur, chef, coach, care giver, cleaner, student, teacher…and the list goes on. Their online habits reflect this diversity and complicate things for advertisers.

Different online channels play different roles in women’s lives, dictating different advertising strategies for marketers. For example, social networking and instant messaging are tapped most frequently for casual interactions. Email retains its status as a powerful online channel in terms of overall scale and ease of content- sharing. FIGURE 13, A DAY IN THE LIFE OF WOMEN ONLINE

(This depicts a typical day in the digital life of women, photographed from focus group materials.) Source: Yahoo / Added Value, Yahoo! Connectonomics (U.S. Study)

When it comes to ticking-off the most entries on the 16 need states list, however, popular lifestyle sites like Shine, online groups and special interest sites rank high. Perhaps counter-intuitively, the anonymity associated with these content sites can lead to deeper emotional connections for women because they provide access to solutions from peers, free of judgment or censure.

Three in ten respondents in the Connectonomics: U.S. study found advertising encountered on content sites to be relevant and engaging.12) Nearly four in ten women said that content sites delivered helpful information about new products and brands.13) Content sites consolidated their role in customer acquisition when almost 45% of study respondents agreed they were a good place to get information about products and brands.14)

Connectonomics sheds light on the best sites to place advertising, sponsor content and even contribute content. The study provides a reliable framework for evaluating need states, who people share with online, how they share, what they share, and how advertising influences the ecosystem, which relates back to the equation: ME = f (N + C + R) Marketing Effectiveness is a function of needs plus channel plus receptivity.

Taking into account all these dimensions, marketers can flight different types of messages, expect certain types of actions and plan for specific levels of ad receptivity. Brand messaging should take consumer needs into consideration and be nuanced accordingly. The framework also provides marketers with a better sense of success metrics. For example, some channels deliver less well on sharing and marketers can optimize expectations for those metrics accordingly.

RECEPTIVITY: CHANNELING SUCCESS All online channels are not created equal. When it comes to advertising impact, some channels deliver greater returns than others due to enhanced receptivity levels. Women in the Connectonomics U.S. study proved three times more receptive to marketing messages delivered via lifestyle, specialty and review sites than other channels. Delving deeper into the reasons behind this phenomenon revealed the reason why: credibility.

Some 42% of survey respondents said women’s lifestyle sites carried reliable information about brands and products; significantly more than the scant 6% of women who said the same about social networking sites. Key learnings from the study suggest specific tactics for matching content and channels such as: women’s lifestyle sites -- gear information toward new product information, buzz and trends; special interest sites -- tailor advertising specifically to the need states of women; review sites -- emphasize product/brand benefits and special bargains/ deals; blogs — target content specific to the interest area; online communities — sponsor editorial content and keep advertising topical; email — nurture relationships with opt-in programs and focus on more graphics, less copy; instant messaging — adopt a quick, light, fun attitude to reflect viewer’s more casual frame of mind; social networking — explore ways to integrate ads into the conversation to avoid feeling disruptive; and Twitter — consider the short shelf life and quick turns for messages, requiring timely, catchy material.

CONCLUSION Given the current growth trajectory of digital media, women are leveraging online communications at an unprecedented rate, already accounting for almost half of global Internet traffic. Despite the egalitarian nature of online media, a gender divide remains, with men more likely to pursue traditionally masculine-oriented content such as online gaming, news and multi-media viewing and women more likely to opt for shopping, email, organizing and personal interest sites. Women worldwide spend half their digital life engaged in connecting activity, satisfying the need for relationship and emotional context.

Women around the world exhibit remarkably consistent needs and behaviors on the web as measured by ~16 discrete factors. Truly, they are electronic “sisters” in every sense of the word, sharing a need for self-improvement/personal growth, mutual sharing/ bonding and being enabled/informed. These need states correspond to online channel selection, where web site attitude and functionality influence audience receptivity, dictating the tone and content of advertising.

Digital media functions as an effective port of entry for advertisers, with women in both developed and emerging countries demonstrating greater receptivity to advertising during every stage of the purchase funnel than men.

Capitalizing on the potential of digital media requires a thorough understanding of women’s need states and motivations, and how those translate across the various online channels. The more closely ad messaging aligns with the media and motivation, the more it will integrate into the digital lives of women all over the world.

FOOTNOTES 1. “Internet Users in the World Distribution by World Regions-2011,” http://www.internetworldstats.com/stats.htm 2. TNS Digital Life 3. TNS Digital Life 4. TNS Digital Life 5. comScore Media Metrix, July 2011 6. comScore Media Metrix, July 2011 7. “Psychology of Male and Female Communicative Activity,” S.A. Vasyura, Spanish Journal of Psychology, May 2008, Volume 11, p289-300. http://www.ncbi.nlm.nih.gov/pubmed/18630669 8. comScore Media Metrix, July 2011 9. comScore Media Metrix, July 2011 10. comScore Media Metrix, April 2010 11. comScore Media Metrix, July 2011 12. Yahoo! Connectonomics Study, Slide 40 13. Yahoo! Connectonomics Study, Slide 41 14. Yahoo! Connectonomics Study, Slide 42

DATA SOURCES comScore Media Metrics 2010 and 2011, TGI (Kantar Media), Global Monitor (Futures Company), other audience data where applicable/available

THE AUTHORS Amy Janis is Senior Manager, Strategic Insights, Yahoo!, United States. Dan Brilot is Research and Development, TNS Digital, United Kingdom. Brian Cooper is Vice President of Technology, TNS North America, United States. Sarah Elliott is Senior Analyst, Strategic Insights, Yahoo!, United States. David ludica is Director, Strategic Insights and Research, Yahoo!, United States. Tony Marlow is Director, Strategic Insights, Yahoo!, United States. TECHNICAL APPENDIX: KEY SOURCE METHODOLOGIES

TNS Digital Life Study Digital Life covers 46 markets, utilizing cutting edge techniques to uncover digital behaviors and expert analysis to understand the attitudes and needs that drive them. Digital Life offers both a lens on the digital world, and the frameworks required to make actionable business decisions within those frameworks. It can be used to drive global strategies or inform local tactics.

Yahoo! Connectonomics Study Yahoo! partnered with the research firm Added Value to conduct a quantitative online survey among almost 5,000 women across the globe, supplementing findings with results from qualitative explorations including over 40 one-to-one interviews and eight focus groups in New York and Colorado. The resulting study provides a new framework for targeting women online, one centered around the needs of today’s woman.

Because connecting online represents the largest share of female Internet time, the study explored the needs associated with their use of the following communication channels and sites (where women could connect with others): social networking sites, Twitter, email, online community groups, instant messenger, blogs, review sites, women’s lifestyle sites (Yahoo! Shine, iVillage, SheKnows, etc.), and special interest sites (CafeMom, BabyCenter, etc.).

PART 2: ONLINE RESEARCH DIMENSION

THE RESEARCHIFICATION OF GAMES ADOPTING A GAME DESIGNER’S APPROACH TO MARKET RESEARCH Peter Harrison

INTRODUCTION In 2010, Nike worked with Wieden and Kennedy to create a game, and something rather special happened. Nike knew that young Londoners found jogging a boring, solitary activity, so they invited a large group of them to play Nike Grid, a running game. They turned London into a game board, divided into zones delineated by post-code. Nike Grid challenged competitors to ‘claim their streets’ by completing as many runs as they could. Additional points and badges were awarded for being the fastest, completing special routes or unlocking ‘secret missions’ like running at night. Players were given a special code that they dialled at designated phone-boxes at the start and finish of their run; this timed them and recorded their scores. Results were published online and shared on Facebook where competitors could gloat and commend each other.

The results? The game inspired 5,000 ”gridders” (the game even created its own terminology!) to run over 12,500 miles, over 31,000 runs in 16 days.1) The campaign also won the 2011 Grand Prize in the Outdoor Planning Awards. The comments on the Facebook group are testament to how they inspired people to participate.

FIGURE 1, SCREEN GRAB OF THE NIKE GRID FACEBOOK WALL

What had been considered a boring, solitary activity had become a social experience energised by competition and recognition. Nike had changed the experience of running.

Games are rapidly taking on a greater role in our everyday lives and culture. The exponential rise of social media and the widespread adoption of advanced mobile technology has introduced new types of game and taken them to incredible new sizes of audience, with games like FarmVille attracting over 80 million players, World of Warcraft, 12 million players, and FourSquare, 6 million. Nike Grid is just one example of how marketing and advertising have increasingly been using games to capture the imagination of their audiences, influence their behaviour and create brand experiences. The popularity of games, together with their ability to engage and educate, has alerted marketers to their potential.

In a trend known as ‘gamification’, defined by Deterdling et al as “the use of game design elements in non-game contexts”2) marketers are now incorporating aspects of games (“game mechanics”) in their campaigns. Classic examples include: points and rewards (loyalty cards); competition (eBay); and finite resource (GroupOn). Although ‘gamification’ is a strong trend in the marketing industry, it is controversial within the game design community, as Margaret Robertson states:15) “Gamification, as it stands, should actually be called pointsification, and is a bad thing because it’s a misleading title for a misunderstood process.” Gamification can make certain activities more motivating, but they fall short of being considered ‘games’ in their own right.

While their use in advertising might be relatively new, games have always been an essential element of human experience. Throughout history they have provided a safe environment for people to experiment, teaching vital life lessons to children and adults alike. Something seems to happen to people when they play games: they develop a sense of flow and engagement and they are motivated to complete challenges to the best of their ability. Games create energy and collaboration. They allow us to explore ourselves by providing the opportunity to experience novel situations.

But why should marketers and advertisers have all the fun? What can researchers learn from games? If we step back a moment to think about it, games have many qualities we seek from market research – to help us learn, experiment, engage, solve challenges, create energy, promote collaboration and learn about people.

This paper will provide a working definition of what a game is. It will explain why the role currently assigned to games by market research misses the broader opportunity that they offer. The paper will show how research games can get us closer to real-life mindsets. It will show what market research can learn from game designers and how we might create great games and research experiences that can enormously benefit clients seeking to understand and predict behaviour.

GAMES: A DEFINITION Jesse Schell is CEO of Schell Games. A renowned conference speaker, and author of “The Art of Game Design”, he defines a game as “a problem solving activity approached with a playful attitude”.4) We find this to be a very useful definition. By placing the emphasis on ‘problem solving’, games move from being limited to activities with no value or purpose outside of themselves, to activities that can also solve real world problems, big or small. The ‘playful’ aspect differentiates games from work – they are activities entered into willingly by players for an intrinsic emotional reward.

This definition suggests that there is a role for games in research, after all what is research if not “a problem solving activity approached with an analytical attitude”? So if research and games already have the same purpose of solving problems, is there anything researchers should know about how we might employ games to the best effect? And what are the benefits of solving problems playfully anyway?

GAMES AS AN INCENTIVE In the main games are seen by the research industry as a means of overcoming declining response rates; as little more than a sophisticated incentive; as a way to get people to give us the answers to the questions we want to ask and to come back for more. As Swahar and Swahar state: “Designing the survey by the same principles that govern games will make the surveys more engaging and thereby facilitate better response rates.”5)

The proliferation of interesting things to do on-line (Facebook, Twitter, e-mail, games, instant messenger, news, etc.) has had an impact on the number of people prepared to spend time completing tear-inducing boring online surveys. Betty Adamou reports that during her time at a sample provider in the UK response rates fell from 10% in 2008 to 2% in 2009.6) At present, this is manageable, but if this trend continues then it will become increasingly difficult to find sample, and even harder to make the argument that it is representative.

Boredom results in a deterioration in the quality of responses – for example ‘flat-lining’ or cursory answers to open ended questions, so the argument goes. If we are not providing a more positive experience for respondents, then they won’t

respond in a meaningful way or they won’t respond at all.7) Games are seen as a helpful answer to this problem, and indeed games can increase engagement in our surveys and communities.

Jon Puleston of GMI has experimented with making surveys more fun by adding game dynamics. Puleston has taken standard question formats and made them more playful, often by framing questions differently or with the clever addition of metaphor. For example, in one question format you are a snowboarder who has to steer quickly through the relevant ‘answer gate’. Another reframes ‘questions’ as ‘challenges’ with a time limit. He has found this especially effective for recall / memory tasks. His gamification of research has recorded a number of benefits:

 ‘Straight-lining’ reduced by 80%;  Neutral scoring cut by 25%;  Respondent experience ratings dramatically increased;  Drop-out rate on some questions reduced from 5% to 1%.8)

Gamifying surveys changes the experience of research for the respondent. This has a range of benefits, namely an improvement in the quantity of data, i.e. fuller qualitative responses and more time spent on questions, which is why Puleston says “gamification … is the most powerful technique we have ever come across in the four years of research we have done to explore how to improve the feedback from online surveys”.9) From a recruitment perspective, gamifying surveys can reduce drop-outs and, in the longer-term, will mean more active respondents, lower recruitment costs and better representation.

But gamifying questions can also change participant’s responses. This happens for a few reasons; often a game will subtly change the answer you are looking for and competition can motivate people to respond differently. There is also the possibility that games can shift the mindset of the respondent, as Puleston says: “entering a game thinking mode can however shift peoples’ mindset somewhat and this can impact on the answers they give.”10) Gamifying surveys can change the way respondents think and the way they answer questions. Whether this leads to ‘better’ or ‘worse’ responses from Puleston’s work is hard to judge, but the differences are certainly observable. Gamifying surveys simply to make them more engaging can introduce effects that we don’t fully understand until the survey has been completed.

Perhaps we need to first think about the mindset we want the participant to be in and then design a research game to create that. This might actually tell us something helpful that traditional approaches miss.

GAMES: A MOOD-ALTERING DRUG A research game we created for the MRS Research 2011 conference led us to a fascinating thought and something of an insight: games have the power to influence our mental state. The starting point for our game was a well-known board game, which we knew would work as a game, and which we modified to have a research purpose.

The game is called “Mopopoly” and is based on the popular board-game “Monopoly”. We designed the game to answer the research objective of understanding more about people’s daily frustrations across a number of categories in the home. For the player the objective is to acquire as many well-being points (the currency of the game) as possible, and to beat the other players by laying claim to the squares on the board, which are not properties (as they would be in Monopoly) but ‘frustrations’. To begin the players assign pre-determined categories of frustration (e.g. laundry, cleaning, cooking, gardening, cars, etc.) to the sets of ‘properties’ around the board. For example, the players might agree to allocate laundry to the most expensive (most frustrating) squares on the board. When they are ready to start the players move around the board in the usual way (i.e. rolling two dice). When they land on a frustration space (a property) they wish to own they have to name a specific frustration / anecdote in relation to that particular category. So for example, for a laundry space they might say “the machine is so noisy it drowns out the TV!” Inability to come up with something means that a player cannot buy or lay claim to the frustration, whilst landing on a frustration space owned by someone else means they have to pay well-being points (i.e. rent) to the owner, and pay double if they can’t come up with their own frustration relating to that category! We altered other game-mechanics such as Chance / Community Chest, to make them more relevant to the new game, with blanks on the cards for players to complete/fill in the reason for the reward/penalty (e.g. “…means you have to cancel your night out. Lose 150 Well-being points). Water Works and the Electricity Company became ‘Hung-over’ and ‘In-laws Visiting’ (a deadly combination!).

The sense of competition and fun felt by players during the game made them forget they were conducting research at all. In fact, the game seemed to create a ‘hot state’, where people were willing to share some incredibly personal anecdotes. In piloting, one very senior member of BrainJuicer staff revealed details about the way he tidies away his children’s underwear, while another talked about the constant and seemingly irrational battle he has with his wife over the size of their laundry basket. At the Research Conference, one player revealed a very particular gardening frustration. He explained

he had just finished raking up the autumn leaves, when someone called round to invite him to the pub. He returned after the pub to find his chickens had walked the leaves around the garden again, creating a terrible mess, and he had to start again. These are the sorts of personal anecdotes that are not only recalled, but also relived and recounted more colourfully and truthfully when in a hot state than in a more neutral frame of mind. Moreover, people’s frustrations were picked up by other players who had experienced a similar issue, promoting further excited discussion and experience swapping. Unconstrained by questions from moderators, it was the rules of the game that actually seemed to provide a loose guide and the freedom for players to express themselves.

FIGURE 2, MOPOPOLY PLAYERS AT THE MRS RESEARCH 2011 CONFERENCE

The game taught us a valuable lesson about the potential that games have for research. Whilst in a ‘hot state’, players’ energy levels were high and they gesticulated a lot whilst talking. Inhibitions were lowered and this meant they shared things with the group that they may not have done otherwise.

Behavioural Economists talk about the hot-cold empathy gap. The hot-cold empathy gap refers to the inability we have whilst in a cold state to predict how we will behave during a hot state. A classic example of this phenomenon is highlighted by research on behaviours related to sexual health, by Dan Ariely and George Lowenstein.11) Respondents answered questions based on their likelihood to have unprotected sex and to conduct other morally ambiguous sexual activities. One set of respondents were in a neutral, un-aroused state (cold state), whereas another group were ‘self-stimulated’ and shown explicit material whilst taking the survey (hot state). Respondents in the hot state condition admitted to being far more open to unprotected sex, a willingness to go further in their efforts to receive gratification and being open to a wider variety of sexual experiences. People in the cold-state were less able to predict their behaviour in a hot state, and this has serious implications for how we learn about people’s behaviour in these situations. Although in this example they did not need to use a game to recreate the experience, the purpose of explicit material was the same – to create a specific mood in the respondent.

Most market research is conducted in a neutral and passive context, and whilst we are in a passive mood we find it hard to empathise with the ‘us’ that’s in a more worked up mood. We don’t even share the same memories in a hot state that we do in a cold state (‘context dependent memory’ is actually quite an old idea in psychology). Mopopoly, whilst not as extreme as Ariely’s experiment(!), did create a hot state amongst the players that put them more in touch with how they feel when they encounter life’s grievances and frustrations. Body-language alone showed us how players got fully absorbed into the moment and were reliving their frustrations with the group. This led us to something of an insight on how research might use games: games possess a unique ability to alter the mood of participants to better reflect or even replicate real-life states of mind. This enables participants to access the appropriate thoughts, feelings, memories and likely courses of action that occur in the moment of interest.

Researchers tend to pride themselves on being able to control for external variables, on asking non-leading questions, ordered in a very particular way, with balanced scales to elicit ‘neutral’ responses and ‘all things being equal’ results. In

many ways the survey is like a science lab – free of any external factors that could contaminate the findings. The inclusion of games in surveys therefore feels instinctively wrong to many researchers, as if it could distract respondents from giving neutral answers. But this is precisely the point and their main benefit. As clean and fair as traditional research approaches seem, they do not recreate the state of mind people are in when making decisions, where any number of distractions are present and we are affected by various moods and feelings.

Research methodologies that can shift respondents from a cold state to a hot state, or from one emotion to another, might actually be more representative of reality and more revealing. Puleston’s work, the Ariely experiment and anecdotal feedback from the Mopopoly game all demonstrate that games can and should do this and can therefore answer some of the problems raised for the industry by Behavioural Economics. Rather than trying to gamify research in a simple bid to make it more engaging, we feel we should be researchifying games; designing games that can help improve research by getting participants closer to the context, mindset and environment of the moment of interest.

ACCOUNTING FOR STATE OF MIND IN RESEARCH “If any of the research techniques used has induced a frame of mind that is not present during the actual consumer experience, it is unlikely to have obtained an accurate picture of what people think.”12) (Philip Graves)

Behavioural Economics has alerted us to inconsistencies between attitude and behaviour; and seeks to explain how and why we often act in ways that are not in our best interest. Daniel Kahneman has adopted the idea of a dual process model from cognitive and social psychology, which he terms ‘System 1’ and ‘System 2’ thinking.13) System 1 thinking is about our intuitive, often unconscious thoughts dominated by feelings and emotions and formed by associations. It is fast, generally ‘felt’ rather than ‘thought’, and helps us make judgements using little energy. System 2 is the rational part of our brain that is responsible for logical thought and reasoning. This is relatively slow to work and extremely energy intensive. Whilst economics is underpinned by an assumption that people make rational decisions that are in their best interests, Behavioural Economics places more focus on the role of System 1 in our decision making and how it can lead to seemingly irrational decisions. Advances in neuroscience over the past ten years have also supported this finding and have shown that often we have made a decision before we are consciously aware of having made it.14) There is a dialogue between the two systems, but in many instances it is System 1 that has the final say!

Whilst both systems are important in our decision making, marketers and researchers often underestimate the role of System 1, especially during decisions like choosing an insurance product, which might seem predominantly rational. Realistically, bringing our System 2 to bear on every aspect of every decision we make would be exhausting and time consuming, and System 1 provides useful shortcuts. Creating research that better understands System 1 responses, rather than just System 2, will help us to gain a better understanding of people’s decision making.

Behavioural Economics also shows that our choices are often susceptible to emotional states, physiological factors, social pressure, local environment and most importantly cognitive shortcuts, as well as the way choices are presented to us. In many cases, System 1 has the upper hand. This makes it very difficult for traditional research to explain and predict behaviour that is context dependent. After all, we cannot account for every different context a decision is made in. Anyone who has ever shopped whilst hungry, bored, angry or hung-over will know that they make vastly different decisions to how their ‘normal’ self would. Experiments by BrainJuicer have shown how the hunger of respondents can have a significant impact on what food to buy; it even changes self-perception (e.g. hungry people are more likely to view themselves as extroverted).15)

Surveys and research communities all exist in virtual environments that are distinct from our everyday lives. Whilst traditional surveys and communities are somewhat limited in their ability to manipulate the experience of the user, adding a game element allows us to transport the participant closer to what they experience when they actually have to make the decision. This is where the most exciting opportunity lies for games – making respondents forget they are doing research. Games provide us with the opportunity to let the respondent re-live the experience under study and access relevant feelings and associations.

CASE STUDY: CREATING EMPATHY To explore this opportunity we have designed games to recreate the mind-sets of people when they are in particular situations. One of these is empathy. Our client, an international pharmaceutical company, wanted to know how people understood and described problems with their digestive system. Rather than asking this as a straight question, we set up a scenario in an online community. In this scenario a friend is complaining of abdominal pains and needs help, the respondent's task was to reach a diagnosis by asking questions. The moderator played the role of the ill friend (Matt or Lisa) and responded in character – keeping the game going. By looking at how respondents played the game, we were

able to gain a strong understanding of the language they used and the way they thought about the symptoms. Some wanted to leap straight to the advice before diagnosing the problem!

FIGURE 3, MATT AND LISA – OUR ILL FRIENDS FROM THE INDIGESTION GAME

For the second part of the game respondents had to recommend to their friend what they should do, their suggestions provided an intriguing range of advice about particular courses of action to take or avoid.

The game aspect of this project did help engage respondents; we saw more of them responding, more often. But this is not the main benefit. By setting up a game where respondents were put in a realistic scenario, we believe the responses they gave were closer to real life and provided a stronger basis for analysis and insight. This meant adopting moderation skills more in line with ethnography than traditional qualitative moderation. It also means that analysis conducted on (virtual) behaviour rather than perceived attitude. This goes some way to responding to Philip Graves’ criticism of research by recreating the frame of mind of the respondent in the relevant situation.

The game helped participants empathise with Matt and Lisa’s problems and relate them to their own. Perhaps because of this they were confident in asking about symptoms and linking these with potential dietary causes and their subsequent medication. Giving the respondents two options of who to help (male or female) was also interesting; respondents were more likely to help people of the same sex, which highlighted the importance of familiarity to the problem - after all giving bad advice is risky!

This example demonstrates that games do not need to be highly technical flash-based pieces of design. Rather, we simply need to adopt an approach that ensures we are creating a relevant experience for respondents.

Betty Adamou is currently working on a more technologically advanced version of the concept with her game “Pimple Crisis”. This game recreates the bathroom setting and gives the respondent the scenario of having a pimple outbreak to deal with. The game mechanics allow the ‘playspondent™’ to select products from around the wash basin to use on his/her face. This is a scalable idea where any room / product range can be programmed. The purpose of this is not just to create a more enjoyable experience for the respondent but also a more life-like representation of the situation; this will get us closer to uncovering the true behaviours and attitudes of respondents.

HELPING RESPONDENTS TO THINK LESS As previously discussed, many of our decisions are guided by System 1 thinking that is intuitive and instructive rather than deliberate. The previous two games helped us understand how to adopt a more System 1 friendly approach to research by creating lifelike experiences that seek to replicate the in-the-moment mindset. Another feature of System 1 and System 2 thinking is that under extreme cognitive load, our ability to process decisions with System 2 wears down and we are more likely to make decisions using System 1. If we are distracted, tired or under time pressure while supermarket shopping, for example, we are less able to think evaluatively about our decisions. Our goal in research therefore should be to find ways to get respondents to think less deliberately and more intuitively when answering questions.

This effect was displayed in a project we did for a multinational manufacturing company. They believed consumers were selecting a rival brand due to the packaging of its product, which was perceived to be more intuitive and to operate on a System 1 basis. We ran a hall-test with two cells, one where respondents had unlimited time to make a decision between the brands and the other where respondents had a time-limited response. We uncovered that the competitor brand was indeed more popular among respondents operating under time-limited (System 1) conditions. The effect was so powerful that although our client’s product was ahead in the unlimited time test (conditions favouring System 2), the competitor

overtook it in the restricted time setting (System 1 conditions). This illustrates our thought that the amount of attention we are able to dedicate to a decision affects the decision that we make.16)

Games might therefore offer us a way to put people under cognitive loads, so forcing them to respond on an instinctive System 1 basis. One example might be a game whose sole objective is to keep the respondent distracted whilst they are answering a survey. On the survey screen are a number of spinning plates on sticks and the respondent needs to pay careful attention to stop them from falling and smashing. At unpredictable times a plate may start to wobble in which case the respondent needs to intervene by ‘spinning’ it around again. This could re-create the level of distraction respondents have whilst shopping. Clearly there is no way to exactly replicate the cognitive load a shopper is under, especially as it differs from person to person and across shopping experiences. But a methodology that helps us understand the effects of cognitive load vs. no cognitive load could provide useful insights into the effectiveness of packaging under different circumstances, for example.

Behavioural Economics has brought to our attention the importance of context on our decision making process. We know our intuitive System 1 often takes precedence over our deliberative System 2, but existing research approaches often neglect this. Games offer us a means to simulate distraction and cause respondents to think less.

THE OTHER SIDE OF THE COIN: USING GAMES TO UNDERSTAND CONSUMER EXPERIENCES So far we have looked at games that recreate our decision-making mindset. We can also use games to understand what shapes this mindset. For example, we might use a game to help us establish where people linger in the supermarket and help us to uncover the unconscious and conscious reasons for this. Wood, Samson, and Harrison17) provide a framework that divides Behavioural Economic thinking into three areas: Individual, Social and Environmental (choice and local environment). To gain a better understanding of how these areas contribute to our behaviour we designed a mass- ethnography game to explore them in more detail.

We recruited participants into one of four groups led by a fictional detective: Miss Marple, Poirot, Columbo and Sherlock Holmes. Each detective had his or her own approach to solving the case that related to our Behavioural Economic framework (environment was split in two – the choice environment and the physical environment):

 Miss Marple - Individual factors: her deep insight into what makes people tick and natural intuition mean Miss Marple is able to identify people’s feelings and curious habits.  Poirot - Social factors: ‘Papa Poirot’ would routinely use his social skills to get close to witnesses, it is this social awareness that makes him a successful detective.  Columbo - Choice factors: A keen eye for detail and a questioning mind make Columbo a natural at understanding the underlying influence choice architecture has on our behaviour.  Sherlock Holmes – Local environment: Completely logical, Holmes is able to dispassionately observe environmental evidence to draw conclusions. "From a drop of water", he writes, "a logician could infer the possibility of an Atlantic or a Niagara without having seen or heard of one or the other."

This was an ethnographic game where the participants played the role of the detective’s assistant; they had to find evidence to help their detective prove their case was the strongest. They were investigating the behavioural cues that make people drink irresponsibly in pubs, bars and clubs. Each detective helped their assistants by providing leads based on their investigative approach, which we wrote in the detective’s tone of voice. To complete the investigation the assistants had to go to a pub (we only recruited people planning on doing so!), Once there, they used the leads from their detective to steer their observations. They then posted their findings on to a community site with their fellow assistants. In a second incarnation of this project, we used Ethos (a tool developed by Everyday Lives) that allows people to capture evidence using their mobile phone’s camera. This gave us even more visual evidence to use and an immediate way of capturing information before post-rationalisation creeps in. Throughout the project all communication with the assistants was in the tone of the detective to add to the fantasy of the game.

This project could equally have been run without the game element. We chose to translate Behavioural Economics into a game to make the topic less daunting for respondents (Behavioural Economics can be quite dry) and to add a sense of play. It was our belief that by making a game of it, especially with the competition between detectives, the respondents would be more engaged. Asking people to remember to do a research project whilst in the pub with friends is difficult. Framing it as a game, however, helps it become an enjoyable activity more in keeping with the spirit of a night out. This is an important consideration for research; for it to be ‘invisible’, it needs to blend in with the type of behaviour it seeks to measure. The responses we got pay tribute to this, the assistants took to the game and provided long and detailed descriptions of their observations. Despite there being some cross-over between the groups, we also found that they were

able to interpret the leads and find Behavioural Economic clues relevant to their detective. Here is a small selection that illustrates some of the contextual factors that impact upon people’s drinking:

 Miss Marple - Individual Factors: “a lot of people knew they’d probably had enough to drink, but after several drinks their inhibitions were much lower / non-existent and so they decided to carry on anyway.”  Poirot – Social Factors: “A lot of us started off by buying rounds of pints/spirits which is probably one of the main reasons that lead to drinking too much – everyone feels they should buy a round and so a lot of booze goes down.”  Columbo – Choice Factors : “Why are soft drinks mostly in the fridge below the level of the bar, so unless you go to order you don't know what they have - I'm female and don't like going up to the bar if I can help it - so I'd like the soft drinks made more obvious.”  Sherlock Holmes – Local Environment: “As they are holding their drinks in their hands, they drink them faster.”

To continue the story of the game into the reporting we also designed the deliverables around a traditional police investigation style dossier (see figure 4). Here is another opportunity for research games: inspiring playful deliverables that imbue the findings with the spirit of the game could also help raise the profile of market research within client organisations by encouraging sharing behaviour!

FIGURE 4, EXAMPLE OF THE REPORT STYLE OF OUR DETECTIVES PROJECT

HOW DO WE CREATE RESEARCH GAMES THAT WORK? “Ultimately, a game designer does not care about games. Games are merely the means to an end. When people play games, they have an experience. It is this experience that the designer cares about. Without the experience, the game is worthless.” (Jesse Schell, “The Art of Game Design”, 2008)

The purpose of using games as a research medium is to create an experience for the respondent. This experience should fulfil one of the following criteria:

 Enhanced engagement with the topic to increase willingness to share and get involved (e.g. Behavioural Detectives / Mopopoly);  Establish a similar cognitive state whilst taking the survey as is experienced during the behaviour under study (e.g. distracted);  Establish a feeling that takes them closer to the behaviour under study (e.g. empathy in the stomach game).

Inevitably, learning from game designers about their approach to creating experiences is useful to us. Jesse Schell in his book “The Art of Game Design” provides a useful framework to use (see figure 5).

FIGURE 5, SCHELL’S ELEMENTAL TETRAD (DESCRIPTIONS ARE OUR OWN)

(Descriptions are our own)

Schell asserts that there are four elements to consider when designing a game. A universal theme should link all these four elements. The universal theme is the point of the game and the experience we are trying to replicate. All four elements should work together to contribute to this theme. Let us consider them each individually:

 Story: The story is an important part of providing the context to the game. How does it apply to the research problem we are trying to solve? For games where we are not trying to recreate a literal version of the decision-making moment, then the story helps to add purpose, a reason for someone to play it.

 Game mechanics: This specifies the rules and structure of the game. What is the goal and how does it recognise success? Do our mechanics allow the respondents to express themselves in a useful way, and are they encouraged to do so? In qualitative games, rules may be broader and slightly fuzzier, although rule-breaking might be an opportunity to learn in itself. Clearly if respondents feel incentivised to lie then there may be a problem with the mechanics!

 Aesthetics: Good aesthetics relate to the universal theme and imbue the game with an enhanced sense of fun and purpose. A focus on the detail and presentation is an important way to make the respondents immerse themselves in the game and get the most out of the experience.

 Technology: This is the board in board games or the computer for computer games. For researchers it might be the survey / community software. How does this affect the experience? Is it seamless and appropriate for the project? In some instances the mobile phone can be an interesting technological add-on as it allows the respondent to capture content on the fly.

By considering all four elements in our design and ensuring the universal theme pervades all four, we can ensure our games meet their objective. Effective games require all four elements to work together to create one powerful experience. This can be a challenge as each element requires distinct skill-sets; consider the vast multi-disciplinary teams required to create modern computer games (artists, writers, programmers, etc.). Whilst research games are not likely to be this complicated, it is important to think beyond just the mechanics and to consider how to construct the story, design the

aesthetics and choose/build the technology. Thinking more broadly about how we design our games will ultimately provide a better respondent experience and responses that are more incisive.

The best place to start when designing a game is to understand the experience that you want to create: what emotions or cognitive states do you want to produce? Then consider any other parameters involved – for example is it a qualitative or quantitative study? Could an ethnographic methodology work? Once you have a concept for the game consider the four aspects of the Tetrad – how could you optimise each of these to create the best game possible? As well as considering the experience, you will also need to consider how the data you get back will be of use as it will often be behavioural rather than attitudinal. Rooting our thinking in the game element first – rather than adding on game elements to research – is likely to provide a stronger basis for the methodology. It is then necessary to play-test the game extensively; involve a range of people to help you determine whether it effectively creates the relevant experience and if it gives you the data you want. Finally return to the Tetrad and consider which parts may be helping or hindering and make necessary adjustments. Then repeat until happy.

Let us look at how we might have developed our Mopopoly game against the principles of the Tetrad. The overall objective of the research game would be to move people into a hot state where we believe they will be able to remember and describe their frustrations better. The universal theme was to create a game play experience similar to that of Mopopoly with a daily frustration twist.

Story: In Monopoly the story of the game is to create a monopoly (actually, the game was originally designed to warn about the dangers of monopolisation!). In our version it had a slight twist. The aim of the game was to avoid daily frustrations and try to impose some of your own on other people. We enhanced the story with the ‘Family Chest’ and ‘Interruption’ cards that mirrored many of the game’s original cards but with a new twist, e.g. ‘Amazon error in your favour collect 200 well-being points’. Money also became ‘well-being points’, as you expend well-being for every frustration that you encounter. One of the nice things about this game is that the players actually partly build the game and the story themselves by naming the spaces and telling their own stories.

Game Mechanics: The mechanics were largely dictated by the rules of the original (although we removed houses and hotels to make it simpler). The main change we made was adding a rule that you have to name a relevant frustration before buying a property, or to avoid double rent. This was obviously key to give us our research output, but in the context of the story made it a stronger game (landing on predefined frustrations would have been boring). One person played the role of ‘banker’, they did not participate in the actual game-play but would pass judgements on whether people’s frustrations were worthy enough to lay claim to the space they were buying (if played in a real research context, the researcher could play this role). When play-testing this at the research conference we came across a problem with the mechanics. The length of the workshop session put a natural time restraint on the game. This had the unintended consequence of making one of the groups play as quickly as they could, resulting in participants spending less time describing their frustrations so that responses elicited by the game lacked the depth of some of those from the other groups. This issue could have been fixed with clearer instructions in the rules for the banker to be stricter on what qualified as an ‘acceptable’ frustration. This example illustrates why it is so important to play-test games in context!

Aesthetics: Getting this right was critical for the game as the original has such an iconic design; we made the game board to closely reflect the original (down to the way it is folded). The story of the game was reinforced by the renaming of the stations (to children coming to play), the utilities (‘hung-over’ and ‘in-laws visit’) and the tax spaces (‘boiler leak’ and ‘power cut’). This provided a tongue-in-cheek twist to the original, whilst complimenting the story.

Technology: We spent a lot of effort producing quality materials so that the game felt like the original. As you can see from the pictures below we made the board from paper-board, similar to the proper game. The property and interruption / family chest cards were also printed on card with a similar thickness to the originals. Due to logistical reasons we did not want to re-create the money, so one of the banker’s roles was to calculate the number of well-being points accrued by the players. Dice were used to enable movement around the board.

FIGURE 6, THE MOPOPOLY RULES

FIGURE 7, PICTURES OF THE MOPOPOLY GAME BOARD AND CARDS

The Tetrad provides a very useful guide for designing games and getting the most out of them. Its strength is in linking all the elements to a central theme, which helps the game to focus on what is important – creating the experience. It also helps with the development of the game as it develops through play-testing The Tetrad provides a useful framework to identify the problems and find the solutions. If the experience is looking a bit flat, it may be because the aesthetics do not effectively support the story, for example. Part of the fun in designing research games then comes in considering how you create the mechanics to capture the information. This is not as hard as you may first think, as nearly every game has some kind of measurement element to it, which is another reason why games can feel like a natural fit for research.

The Tetrad allows us to focus on the experience we want our research game to create. This is essential if our results are to reflect people’s heat of the moment decisions. So we start with identifying the experience and then align the four elements with it, the game grows from there. The game mechanics provide the means for participants to engage with the game, so this needs to work seamlessly with the other elements to create a singular experience. This game-first approach will allow us to design more creative and effective research games than a focus that limits itself to gamifying research

questions. This is why we call our approach the ‘researchification of games’ and why we believe it will give us more relevant and useful data.

CONCLUSION Games create experiences and this is why they are a useful tool for the market research industry. They generally are considered as a means of improving the respondent experience, leading to fuller answers, fewer dropouts and higher response rates. Whilst it is hard to disagree with the positive effects games have on respondent experience, some voices in the industry are raising questions about how they affect the mindset and subsequent responses of the respondent. Whereas this is generally perceived as a problem, we see it as an opportunity to edge the respondent closer to the mental state of interest. This is the real benefit and opportunity offered by research games: to create context that is absent from traditional research methods, to provide responses that are rooted in the right mindset and therefore have greater real- world applicability.

Market research, especially quantitative, interprets respondents’ understanding of their past and future selves. The accepted wisdom has always been that the best way to elicit accurate responses is by asking direct questions in a context that does not disturb or distract the respondent from their rational thinking process. The problem with this approach is that we are not asking their past or future selves but their present selves. And these ‘present selves’ are by definition taking a survey in a mindset that is not representative of the mindset we are interested in. The key insight for market research from Behavioural Economics is that humans are inconsistent, that our decision making is dependent on mood, physiological state and cognitive resources (all of which are influenced by any number of other factors). The market research industry is pretty poor at taking this into account, particularly in our quantitative approaches.

Games offer researchers a way to change the mindset of the respondent, to put them into the same mental space as they are when making a particular decision. This goes against the perceived wisdom that we should create neutral experiences to get scientifically accurate results. Often this will mean we are not asking questions so much as setting challenges and interpreting behaviour – like in the indigestion game where we draw conclusions from the questions they asked. We have also discussed how we could use games to introduce cognitive load on a respondent to simulate making decisions when stressed or distracted.

This paper has highlighted that research games do not need to be highly complicated or flash-based, or anything on a scale approaching the games industry. When designing a research game the main consideration is the experience we are giving the respondent – what purpose does it serve? We can learn from game designers, such as Jesse Schell, whose Elemental Tetrad helps position experience at the centre of what we create.

This paper has scratched the surface of what can be achieved with research games, but has laid out the philosophical reasons why we believe research games are not just beneficial but are actually much needed by the industry. Researchification of games is an ambitious approach that overturns some of the long-held assumptions of the industry and responds to the new challenges coming from the social sciences. It is an exciting concept with plenty of scope for further exploration and offers market research the potential to get closer to the real lives of respondents.

FOOTNOTES 1. http://awards.akqa.com/awards2011/NYF/Nike_Grid/default.html 2. Deterdling, Dixon, Khalen, Nacke, “From Game Design Elements to Gamefulness: Defining “Gamification”, 2011 3. Margaret Robertson, ‘Can’t Play, Won’t Play’, http://www.hideandseek.net/2010/10/06/cant-play-wont-play/ 2010 4. Jesse Schell, ‘The Art of Game Design’, 2008 5. Gayathri Swahar, Janardhan Swahar, “Designing Innovation: Maximizing Online Respondent Engagement Through A Game-Way Research Design” ESOMAR Innovate, Barcelona, November 2010 6. Betty Adamou, ‘The Future of Research Through Gaming’, 2011 7. Mike Cooke, “The Engagement Agenda”. Association for Survey Computing (ASC): 'Pizzazz in Research: Renewing the Rules of Engagement', 2011 8. Jon Puleston, ‘So online research what’s new?’, 2011 9. Jon Puleston, “The Game Experiments”, 2011 10. Jon Puleston, “The Game Experiments”, 2011 11. Dan Ariely and George Lowenstein (The Heat of the Moment: The Effect of Sexual Arousal on Sexual Decision Making, 2006, Journal of Behavioral Decision Making) 12. Philip Graves, “Consumer.ology: The market research myth, the truth about consumers and the psychology of shopping”, 2010

13. Kahneman and Frederick. “Representativeness revisited: Attribute substitution in intuitive judgment”. In T. Gilovich, D. Griffin, and D. Kahneman, “Heuristics of intuitive judgment: Extensions and applications” (pp. 49-81). New York: Cambridge University Press. 2002 14. For an account of the method, please read Haynes, J and Rees, G Decoding mental states from brain activity in humans. Nature Reviews Neuroscience (Volume 7):523-34 (2006). For details of the experiment itself, please read Soon, CS, Brass, M, Heinze, HJ, and Haynes, JD Unconscious determinants of free decisions in the human brain. Nature Neuroscience (Volume 11): 543-545 (2008) 15. Wood, Samson, Harrison, Behaving Economically With the Truth, 2011 16. For a more detailed explanation of the project see “Behaving Economically with the Truth”, Wood, Samson, Harrison, 2011 17. Wood, Samson, Harrison, Behaving Economically With the Truth, 2011

BIBLIOGRAPHY Adamou, Betty Adamou. ‘The Future of Research Through Gaming’, 2011 Ariely, Dan. ‘Predictably Irrational: The hidden forces that shape our decisions’, Harper, 2008 Ariely, Dan and George Lowenstein “The Heat of the Moment: The Effect of Sexual Arousal on Sexual Decision Making”, Journal of Behavioral Decision Making, 2006 Chatfield, Tom. “Fun Inc: Why games are the 21st Century’s most serious business”, Virgin Books, 2010 Cooke, Mike. “The Engagement Agenda”. Association for Survey Computing (ASC): 'Pizzazz in Research: Renewing the Rules of Engagement', 2011 Deterdling, Dixon, Khalen, Nacke, “From Game Design Elements to Gamefulness: Defining “Gamification””, 2011 Douglas, Graeme. “Young Runners Reclaim the Streets with Nike GRID”, Campaign Magazine, April 2010 Graves, Philip. “Consumer.ology: The market research myth, the truth about consumers and the psychology of shopping”, Nicholas Brealey Publishing, 2010 Kahneman, Daniel. “Maps of Bounded Rationality”, Nobel Prize for Economics Speech, 2002 Kahneman and Frederick. “Representativeness revisited: Attribute substitution in intuitive judgment”. In T. Gilovich, D. Griffin, and D. Kahneman, “Heuristics of intuitive judgment: Extensions and applications” (pp. 49-81). New York: Cambridge University Press. 2002 Kay, John. “Obliquity: why our goals are best achieved indirectly”, Profile Books, 2010 Koster, Raph. “A Theory of Fun for Game Design”, Paraglyph Press, 2005 McGonigal, Jane. “Reality is Broken”, Jonathan Cape, 2011 Puleston, Jon. "So online research, what’s new?", 2011 Puleston, Jon. “The Game Experiments”, 2011 Schell, Jesse. ‘The Art of Game Design’, Morgan Kaufmann, 2008 Swahar, Gayathri and Janardhan Swahar, “Designing Innovation: Maximizing Online Respondent Engagement Through A Game-Way Research Design” ESOMAR Innovate, Barcelona, November 2010 Wood, Samson, Harrison, “Behaving Economically With the Truth”, 2011 http://www.hideandseek.net/home/blog/ www.facebook.com/nikegrid http://awards.akqa.com/awards2011/NYF/Nike_Grid/default.html

THE AUTHOR Peter Harrison is Creative Director, BrainJuicer Labs, United Kingdom.

HOW FAR IS TOO FAR? TRADITIONAL, FLASH AND GAMIFICATION INTERFACES, AND IMPLICATIONS FOR THE FUTURE OF MARKET RESEARCH ONLINE SURVEY DESIGN Bernie Malinoff • Jon Puleston

INTRODUCTION

Purpose of this paper Technology development and adoption is outpacing our understanding of how and when to optimize the use of different onscreen survey options. We are also witnessing an increasing “digital divide” between research firms based on their relative adoption of “traditional”, “flash/rich media” and now “gamification” question types.

As such, we continue to witness a paradigm shift in how technology is ahead of market research industry best practices. How should one choose a Traditional, Rich Media/Flash or Gamification online survey design? Should we take the shiniest ball in the bucket, or are there occasionally times when the ‘rusty’ ball is better to use?

The inbound premise of this paper is that we can't slow down the rate of adoption, or the rate of technological advancement. Nor should we. But we should keep pace as an industry in understanding the new tools we're using – from a point of view of how each of these can impact data consistency and user experience.

Previous research led by element54 “Sexy Questions, Dangerous Results?” (2009 with ResearchNow) and “Eyes Don’t Lie” (2010 with MBA Recherche/UX Research) guides this latest research initiative “How Far is Too Far?”

Sample The sample contained 1,506 nationally representative Canadian adults 18+ drawn from GMI sample split into four cells:

Cell 1: 294 adults who only answered the standard format questions Cell 2: 303 adults who only answered the flash format questions Cell 3: 303 adults who only answered the game format questions Cell 4: 606 adults who answered randomly one of the variants of each question

All four cells were quota balanced by age and gender and exactly weighted before analysis. (See table 1.)

TABLE 1 Total Standard Flash Game Combined 1506 294 303 303 606 Male 49% 49% 49% 49% 49% Female 51% 51% 51% 51% 51%

18-24 years 21% 21% 21% 21% 21% 25-34 years 26% 26% 26% 26% 26% 35-44 years 17% 17% 17% 17% 17% 45-54 years 19% 19% 19% 19% 19% 55-64 years 11% 11% 10% 10% 11% 65+ years 6% 6% 6% 6% 6% Base: All respondents

Fieldwork The fieldwork for all four cells was conducted using simultaneous batches of survey invitations sent out over a five day period from the 4 – 9 August to ensure a good balance of week day and weekend respondents.

Methodology The surveys were created using GMI’s proprietary Qstudio survey design software.

Four versions of the test survey were created and were made up of a cross selection of randomly chosen questions on general interest topics, ranging from politics, media consumption, sport and the environment.

The standard version: The standard question format of the survey used only standard check selection boxes/grids.

The flash version: This used a combination of standard flash format questions including sliders, animated scrolling matrix questions and word picker questions.

The game version: This version shared some of the same flash question formats used in the flash version of the survey but also included a range of more creative flash question formats including flag drag and drops and other more visually decorated questions. It also included two basic guessing game questions.

At the end of all three surveys respondents were then asked to rate the survey on a range of measures and were then invited to make comments about what they liked or disliked about the survey.

The combined version: This version included every question variant used in the above three surveys. On each question respondents were randomly shown one of the three variants, so each respondent would have randomly answered a combination.

At the end of this survey, in addition to giving their general rating, respondents were asked to rate each of the individual question formats they had answered.

APPROACH TO ANALYSIS Consistent with our previous R&D (“Sexy Questions, Dangerous Results” 2009, and “Eyes Don’t Lie” 2010), we believe the survey interface impact can be explored from two important perspectives … Respondent Experience and Data Consistency.

SECTION 1: THE RESPONDENT EXPERIENCE The result of this experiment demonstrates the very strong relationship between the entertainment value of the questions and the willingness of respondents to invest time and effort in thinking about the answers they give to the question. Figure 1 illustrates an aggregate score from all the questions asked in this survey.

FIGURE 1, AVERAGE INDEX COMPARISONS: FUN VS. ANSWER TIME VS. STRAIGHTLINING

We consistently found that the higher the respondents fun rating the longer they gave to answer time; it was also seen that the level of straight-lining was lower.

From a consumer point of view, it is as much a piece of entertainment as anything. Most people do surveys in their spare time purely as “bit of fun”, and as with any form of entertainment, the more entertaining it is, the more time you are prepared to invest in consuming it. This, perhaps, is at the heart of how we can evolve as an industry from Basic, to Flash to Game-based forms of survey interaction.

To underline this point are some specific examples of question by question comparisons from this survey experiment.

Example 1: Attitude statement rating In this first example you can see the three different ways we experimented with asking this rating question. The first is the standard way of doing this using a conventional grid. The second format used sliders and a design to look like a mixing desk, and with the third we used facial expressions to reflect the different rating scores.

The standard version: This employed conventional grids and whilst considered easy and simple to answer the response to the question was muted. “OK” was the key word used. (See figure 2.)

FIGURE 2, HOW MUCH DO YOU ENJOY DOING EACH OF THE FOLLOWING ACTIVITIES?

Flash version: This format used a set of vertical sliders which mimicked the style of a mixing desk. This was seen as just as easy and simple to use as the conventional grid. It was also seen as fun and well designed. Respondents spent the most time answering this question which we investigated through some basic usability testing - watching people answer the question. It appeared to encourage respondents to tweak and adjust their answers in a way they did not bother doing with the conventional grid. This was reflected in a reduced level of straight lining to under 1%. (See figure 3.)

FIGURE 3, HOW MUCH DO YOU ENJOY DOING EACH OF THE FOLLOWING ACTIVITIES?

The game version: This, instead of using a conventional 7 point range, used a set of facial expressions instead. It was seen as just as simple and easy to use as conventional grids, but a lot more fun - which in turn encouraged respondents to think a bit more about their answers. This resulted in zero levels of measurable straight lining. (See figures 4 and 5.) FIGURE 4, HOW MUCH DO YOU ENJOY DOING EACH OF THE FOLLOWING ACTIVITIES?

FIGURE 5, KEY WORDS USED TO DESCRIBE THE ACTIVITIES QUESTION

Standard Mixing desk Faces Answer time (s) 24 32 29 Straight-lining 3.8% 0.4% 0.0%

Example 2: General agreement statements

The second version contrasted two different ways of asking a conventional grid question.

The standard grid: respondents (see figure 6.)

FIGURE 6, HOW MUCH DO YOU AGREE OR DISAGREE WITH THESE STATEMENTS?

The flash version: This used scrolling matrix questions. This time using visual thumbs up and down icons; as each question is answered the next question automatically scrolls. This approach was viewed as less boring, but not seen to be as much fun as the facial expression version above. However it still resulted in 50% less straight lining than the conventional grid variant. (See figure 7.)

FIGURE 7, HOW MUCH DO YOU AGREE OR DISAGREE WITH THESE STATEMENTS?

Game version: Respondents were asked to sort the options into groups in a very simple visual format. This was seen to be an inherently more interesting way of answering this question by respondents, helping them differentiate between choices, again resulting in half the level of straight lining. (See figure 8.)

FIGURE 8

FIGURE 9, AGREEMENT QUESTION: KEY WORDS

Standard Flash Game Answer time (s) 74 78 93 Straight-lining 10.0% 4.6% 4.9%

Example 3: Political awareness question Standard: The reaction to the standard grid version of this question was that it was simple and easy but also dull and boring. (See figure 10a and 10b.)

FIGURE 10A, HOW FAMILIAR ARE YOU WITH EACH OF THE FOLLOWING POLITICANS?

FIGURE 10B, HOW CONFIDENT ARE YOU WITH EACH OF THE FOLLOWING POLITICANS AS A LEADER OF CANADA?

Flash format: In the flash version the option choices automatically scrolled after each selection was made and images of the politicians were used. This was slightly faster than the standard grid format question. Respondent reaction was that they thought it was well designed, more fun and enjoyable and less dull, which resulted in 50% less straight lining. (See figure 11a and 10b.)

FIGURE 11A, HOW FAMILIAR ARE YOU WITH EACH OF THE FOLLOWING POLITICANS?

FIGURE 11B, HOW CONFIDENT ARE YOU IN EACH OF THE FOLLOWING POLITICANS AS A LEADER OF CANADA?

Game version: In the first part of the game version of these questions, instead of saying whether or not they were aware of each politician, they had to bet on whether people were politicians or not - and the more confident they were the more they could bet. This really sparked respondents’ imagination and as you can see, proved almost universally popular and encouraged respondents to spend more than twice as much time thinking about their answers. It also eradicated any observable levels of straight lining. (See figures 12a, 12b, and 13.)

FIGURE 12A, DO YOU THINK THIS PERSON IS A POLITICIAN?

FIGURE 12B, HOW MANY DID YOU CORRECTLY GUESS?

FIGURE 13, KEY WORDS USED TO DESCRIBE THE POLITICAL QUESTION

Standard Flash Game Answer time (s) 21 19 50 Straight-lining 9.3% 5.6% 0.0%

Example 4: Brand awareness measurement

Standard: A basic grid. The main descriptive words used to describe this question were boring and repetitive (see figure 14).

FIGURE 14, WHICH OF THESE BRANDS DO YOU RECALL SEEING ADVERTISING FOR ON TV?

Flash: This used brand logos that appeared on a TV screen and scrolled automatically one by one as each question was answered. This turned the question in the mind’s eye of the respondents from being boring and repetitive to be fun and interesting. Consideration time increased and straight lining was marginally reduced (see figure 15).

FIGURE 15, WHICH OF THESE BRANDS DO YOU RECALL SEEING ADVERTISING FOR ON TV?

The game version: This was essentially the same basic format as the flash version but a game mechanic was added where respondents had to predict whether they thought a lot of other people had seen the advertising for each brand. Now, it must be noted that this was a very simplistic game mechanic and not one that you would necessary want to use in a real life situation, however it was chosen to purely demonstrate the effect it could have. This almost tripled the enjoyment and fun scores, was seen to encourage respondents to spend 40% more time considering their answers and reduced measurable straight lining by a third. (See figures 16 and 17.)

FIGURE 16, WHICH OF THESE BRANDS HAVE YOU SEEN ADVERTISED ON TV LAST YEAR?

FIGURE 17, KEY WORDS TO DESCRIBE THE BRAND AWARENESS QUESTION

Standard Flash Game Answer time (s) 56 63 74 Straight-lining 21% 19% 14%

Example 5: Advertising awareness question For this question respondents were asked to pick out types of media they used for six different information sources which was quite a repetitive task.

Standard questions: This was asked in two ways; one group was asked to openly pick media choices and the second group was prompted to pick first, second and third choices. The vast majority of respondents found both these formats boring and repetitive. (See figure 18a and 18b.)

FIGURE 18A

FIGURE 18B

Flash version: In this version images were used to visually describe each media choice. Respondents were allowed to make open choices and when selections were made they were stamped with an award icon. This was seen to be significantly more fun, better designed and far less boring for respondents than the conventional grids. This was one question instance where respondents actually answered this question faster than the grid version. Employing basic usability testing we discovered that this was more to do with the difficulty in reading the option choices in the grid version of the question. The level of measurable straight lining was reduced by more than 50% (see figure 19).

FIGURE 19, THE INFORMATION SOURCES YOU THINK ARE EASIEST TO USE

The game version: In this version respondents were asked to give out awards to the best media for each information source by dragging and dropping first, second, and third prize icons onto the media shown. Again this method was seen to be more fun and really encouraged respondents to think about their answers, on average each spending over two minutes answering this question compared to 83 seconds for the basic grid version. Straight lining was reduced by 55%. (See figures 20 and 21.)

FIGURE 20, YOUR AWARD FOR THE INFORMATION SOURCES EASIEST TO USE

FIGURE 21, INFORMATION SOURCES QUESTION: KEY WORDS

Standard open Standard 1st 2nd 3rd Flash open Game 1st 2nd 3rd Answer time (s) 79 83 70 124 Straight-lining 10.3% 15%* 4.5% 6.7%* *Note the differences here to open version straight lining was calculated on 1st choice TABLE 2, CONSUMER REACTIONS TO THE WHOLE SURVEY EXPERIENCE

Completion rates Standard 94% Flash* 92% Game* 92% Average (median) completion time in minutes Standard 11.7 Flash 12.3 Game 14.5

As this was a relatively short and interesting basic survey there is little to compare in the basic completion rates, but the flash designed and game format survey saw consumer enjoyment scores being significantly higher. (See figure 22.)

FIGURE 22, SURVEY OVERALL ENJOYMENT SCORES (OUT OF 10)

The open ended feedback was also significantly more positive towards the flash and game variants (figure 23).

FIGURE 23, RESPONDENT OPEN ENDED FEEDBACK ABOUT EACH SURVEY VARIANT COUNT OF KEY WORDS USED

SECTION 2: DATA CONSISTENCY In our previous research, “Sexy Questions, Dangerous Results?”, we had concluded there were wide-ranging differences across survey interface applications (Radio Buttons produced different results vs. Sliders, etc.).

Our hypothesis in conducting this latest round of research was that differences would exist, since the reality of changing the survey interface was not to specifically arrive at how to optimize data consistency, but rather to achieve a balanced understanding of how and where the “wins” would be that can advance a constructive view of moving a critical mass of surveys forward with industry alignment.

We recognize that some behavioral questions are more likely to evoke fewer differences than others. When we asked teeth brushing frequency, we learned that using visuals and more creative approaches to question design significantly improved the consumer experience, and perhaps this is a good example of how you can engage respondents at the start of a survey. In this case, the response patterns were fairly consistent across survey interfaces (see figures 24a – 24c, and 25).

FIGURE 24A, HOW MANY TIMES PER DAY DO YOU TYPICALLY BRUSH YOUR TEETH?

FIGURE 24B, HOW OFTEN DO YOU BRUSH YOUR TEETH?

FIGURE 24C, HOW OFTEN DO YOU BRUSH YOUR TEETH?

FIGURE 25, TEETH BRUSHING FREQUENCY

Changes caused by differences in the selection process A basic example of this is the commonly observed issue of using sliders compared to standard buttons selection. As figure 25 demonstrates, when asked to rate their level of Optimism on a scale of 1-10 (a more attitudinal question), we see the centrally positioned slider format used in the game version of the survey has lower centre scoring which is a commonly seen effect, resulting from the fact that respondents tend to move it left or right but are reluctant to keep it where it is. You could see this as an advantage if you were wishing to pull out differences between option choices, but in this case as there is only one thing being measured you could argue that this manipulated the result adversely. (See figures 26a – 26c, 27).

FIGURE 26A, USING THE SCALE BELOW, PLEASE TELL US IF YOU CONSIDER YOURSELF TO BE MORE OF AN OPTIMIST OR A PESSIMIST

FIGURE 26B, USING THE SCALE BELOW, PLEASE TELL US IF YOU CONSIDER YOURSELF TO BE MORE OF AN OPTIMIST OR A PESSIMIST

FIGURE 26C, DO YOU CONSIDER YOURSELF TO BE MORE OF AN OPTIMIST OR A PESSIMIST

FIGURE 27, LEVEL OF OPTIMISM

Another example of how the data changed based on question interface can be found in the question of “Activity Enjoyment”. We asked a series of rating statements around types of activities using a scale of 1-10 to ask how much they enjoyed each.

As figure 28 demonstrates, the facial emotion format used in the game version of this survey shows that respondents really liked it (fun score increased from 20% to 56%), so on that level this is probably one of the most successful question formats. However it did manipulate the score resulting in an overall more positive rating and 30% more clicks on the 5/7 option. This peaking effect was reduced when a less clown-like female face was used (figure 29).

FIGURE 28, HOW MUCH DO YOU ENJOY DOING EACH OF THE FOLLOWING ACTIVITIES: RELAXING ON THE WEEKEND

FIGURE 29, HOW MUCH DO YOU ENJOY DOING EACH OF THE FOLLOWING ACTIVITIES: DISCUSSING POLITICS

FIGURE 30, ACTIVITY ENJOYMENT

The overall differences could be explained by the fact that what respondents were evaluating was a range of basic domestic activities and the worst of which was washing up and so this will not necessarily provoke strongly negative feelings that the most angry first option image suggests, whereas when using a number scale the scoring appears to be used in a more relative way. More people picked 1 for washing up as it was the worst task in the list comparatively. If the list included an even worse task the result might have been different. You could argue that the facial expression range offers a more stable set of answers less influenced by relative factors.

Another interesting question format which was used over two different questions was a ‘Flash drag and flagging’ approach, where consumers could drag a flag to their chosen answer. The answers to this format were compared to the question answered in more conventional ways - as seen in figures 31 - 34 - which seemed to generate quite consistent data, however with a notable peak in respondents selecting 8 and further, a shortfall in the number of people selecting the 9 score which could be something to do with the psychology of the 10 point rating process.

FIGURE 31

FIGURE 32

FIGURE 33

FIGURE 34

This flag drag and drop approach in many respects was an extremely successful way of asking this question as it combined a forced level of ranking of the opinion and produced a smoother distribution of data whilst completely eradicating straight-lining. But, by direct comparison to the standard rating process, it produced lower 1 and 10 selections, although this can be explained by the fact that these were the limit points.

However when this same question format was used to record drinking instances later on within a survey, this meant each option could be dragged more than once. The extra work involved in dragging and dropping resulted in 25% less drinking instances being marked.

FIGURE 35

FIGURE 36

FIGURE 37

FIGURE 38

Changes caused by the fun factor The other observable factor was that if the question was more fun to answer it could encourage an all-round more enthusiastic approach to answering the question and result in considerably more click activity.

See figures 39 - 41. Respondents were asked to click on various media and by switching to a more fun visual approach - where when respondents made their selection prize rosettes appear - and which encouraged a 50% increase in click activity.

FIGURE 39

FIGURE 40, THE INFORMATION SOURCES YOU THINK ARE EASIEST TO USE

FIGURE 41, % CLICKED ON EACH INFO SOURCE

When we asked people to count the number of energy saving light bulbs in their house, and prompted them with this message - to turn it into more of a fun game - it resulted in two extra light bulbs being reported than when we had simply asked them to make a guess (figure 42).

FIGURE 42

The image effect There can also be major observable differences in answers when images are used to prompt answers. Images can trigger the memory more effectively than words and as a result increase awareness.

For example awareness of sports teams increased by 25% when we employed images in this example question (figures 43 – 44).

FIGURE 43

FIGURE 44

And awareness of brands advertised on TV increased by 27% when logos were used. So too was there an increase in overclaim for certain brands which had definitely not advertised on TV, which finds an increase from 7% to 11%, a net 4% increase.

FIGURE 45

FIGURE 46

The game effects In this experiment we employed two guessing game techniques. In the first one we asked respondents to guess if someone was a politician or not and to gamble money to back their choice. We found here a very close broad correlation with basic awareness style questions and gambling choices. In fact you could argue that the gambling data in differential terms was more accurate as respondents had something at stake in getting the answer right or wrong.

FIGURE 47

FIGURE 48

On the other hand when we employed a guessing game mechanic in the TV advertising awareness question where instead of two binary choices they had to guess as to whether a lot of other respondents would be aware of each brand or not. This significantly reduced the non awareness selection choice and doubled the awareness scores of non-advertised brands and the assumption is that the respondents were wrapped up in the game.

FIGURE 49

This is a good example of how games, when not thoughtfully applied in a survey, can have a quite a disruptive influence on the data. The objective of the game needs to re-enforce the objective of the question for it to work successfully.

SUMMARY What this experiment has essentially explored are some different and more creative ways of asking conventional survey questions, with the goal of fostering industry dialogue and best practices about when and how to optimize survey creation and evolution.

We have estimated that well over 90% of global surveys are done using a “basic” interface (from start to finish). Even providers of Flash based software have inconsistent current day practices about when and how clients should be using different interfaces.

We have learned that there is much to be gained by taking a ‘small’ leap from Basic to Flash, for at least part of the consumer experience. Doing this will greatly enhance survey enjoyment and improve the quality and confidence in our data. However, we recognize that certain types of studies require parallel testing and/or other due diligence before

migrating the interface. Based on our conversations with several Top Global firms, we understand the spirit to do this is alive and well.

However, for large global market research firms we appreciate there are two road blocks to greater adoption and implementation:

1. Scalability – for most large firms, programming is a black box operational cost centre, and in some cases is off-shored for cost efficiencies. For these firms, a case needs to be made for how one can improve the respondent experience without incurring additional programming costs by “thinking” about programming.

2. Norms – firms which maintain normative databases (concept tests, ad tests, tracking, etc.) are mindful of how a survey interface changes from “how it’s always been done” which will impact the normative database, much like the concerns of having migrated from Telephone to Online data collection.

What our research shows is that design and creative questioning can have an impact on the user experience and quality of the data. So whether the interface is Basic, Flash or Game-based, we have an end-goal of making questions more entertaining for consumers and this is a mindset that can be applied to nearly any survey across any channel of data collection.

ACKNOWLEDGEMENT The author wishes to thank GMI Interactive, and especially Jon Puleston for providing survey programming, hosting, sample and data tabulation for this study. This experiment would not have been possible without their generosity, and like- minded commitment to advancing industry dialogue through experimentation and constructive discussion.

THE AUTHORS Bernie Malinoff is President, element54, Canada.

Jon Puleston is Vice-President, GMI Interactive, United Kingdom.

FACIAL IMAGING THE NEW FACE OF ONLINE SURVEY RESEARCH Alastair Gordon • David McCallum • Matteo Sorci• Tim Llewellyn

INTRODUCTION Neuroscience, biometrics, and facial imaging are "hot topics" at the moment in our industry. Much of the discussion is, however, posed as an alternative to and indeed as a replacement for conventional survey research. At best, an online survey is seen by the proponents of these "science based" methods as being a secondary add on to parallel neuroscience studies. The task of survey researchers is to adapt their studies to fit around the needs of the 'superior' new age methods.

In this paper we take issue with this proposition. Online surveys have a huge and continuing role to play in Market Research, and we take the view that it is the task of 'brain researchers' to adapt their methods to integrate and work with surveys, not the other way around.

Our central thesis is that facial imaging offers a means to accurately record emotional response with few or indeed no questions needing to be asked. Since, for many survey purposes (ad testing, reaction to packs, brand image assessment etc.), gauging emotional response comprises a large part of the questionnaire, this means research practitioners can save the bulk of the survey for rational questions (usage patterns, location, demographics, etc.). The survey would be both shorter, and provide more accurate measures of emotional response. Over the last year Gordon & McCallum and the Swiss technology company nViso SA have been engaged in testing this idea using a method called Facial Imaging.

Facial imaging is based around the well-established model that shows that micro-expressions on our face can accurately reflect emotions like happiness, fear, and surprise. Our system has added to currently existing methods a highly intelligent software system that automates the recognition of the most transitory of micro-expressions and combined that with the ability to collect the data using very basic webcams.

Essentially, this means that if people can be persuaded to look at a computer screen and turn on their webcam, their emotional response to whatever they are watching can be gauged. In this paper we will show how this system works, demonstrate why we believe it is a more robust method for the collection of emotional data than traditional question based methods, and illustrate using case studies from several countries the kinds of outputs possible from this kind of advanced research.

We conclude that facial imaging has considerable potential to enhance online research practices. In particular we hope to illustrate:

 How passive facial imaging techniques can shorten and enhance questionnaires.  The system's positive impact on the speed and richness of feedback possible.  The ability to cut through the debate on the use of affective and metaphor based scales.  How the type of measurement possible dramatically increases the granularity and amount of information it is possible to collect in a survey, offering new opportunities for advanced modeling and analytics.  An ability to compare and contrast many very similar stimuli without confusing the respondent.  The impact of webcam interviewing on quality control, and the opportunity to refocus 'rational' questions in the questionnaire.

BACKGROUND TO FACIAL IMAGING RESEARCH Facial Imaging is based on the work of Dr Paul Ekman (of 'Lie to Me' fame) and others, which shows that micro- expressions on our face accurately reflect emotions like happiness, surprise, disgust and fear. Moreover, these fleeting expressions are both innate (we are born with them) and expressed in the same way across cultures and ethnicities. Measuring response to facial expressions is not new in Market Research; companies in the United States and Europe have

been doing it for some time, either using indirect measures (getting people to associate stimuli with faces with known emotional connotations), or by having expert coders manually encode videos of people responding to marketing material. As noted above, Facial Imaging enhances these methods by automating the recognition of facial micro-expressions, and adding in the ability to collect the data using remote webcams.

Since the field of Facial Imaging is a new one, and less well known than E.E.G. and similar neuroscience methodologies, it is worthwhile to spend a little time outlining the development and theoretical underpinnings of this methodology.

FACIAL EXPRESSIONS - HISTORY AND ACADEMIC BACKGROUND

Charles Darwin As early as 1872, Charles Darwin postulated that facial expressions were universal, across cultures, ethnicities, and even species. Facial expressions were essentially behaviours that evolved as a mechanism of communication. They were, he claimed both Innate: expressions are not learnt; and Universal: some expressions are the same in all humans and animals. Notably in this regard, David Matsumoto's work comparing expressions on winning and losing athletes in the 2004 Olympics showed no difference between those at the Summer Olympic Games and those, blind from birth, competing at Paralympic Games, implying that expressions of happiness in victory and sadness in defeat for example are 'hard-wired' and not learnt from observing others' behaviour and expressions.

Paul Ekman et al However, the most prominent work on Facial Expressions is that conducted by Dr. Paul Ekman. Dr. Ekman, ironically, originally set out to disprove Darwin's theory of universality but, in the end, concluded that to all intents and purposes Darwin was right. Ekman's work initially postulated the universality of six emotions (seven if we include neutral), being: happiness, surprise, fear, anger, disgust, and sadness (+neutral). In 1978 Ekman and Friesen developed the Facial Action Coding System (FACS) to taxonomize every human facial expression. FACS measurement units are called “Action Units” (AUs) representing the muscular activity of the face. Currently FACS is the leading global standard for facial expressions analysis.

Sorci et al Work undertaken at the Swiss Federal Institute of Technology (EPFL) over the past decade has resulted in Ekman's work being translated into a machine-learning environment, as opposed to the 'manual' use of human expert facial coders. In essence, the work of Dr. Matteo Sorci et al has developed an Artificial intelligence which precisely detects facial muscles. These Facial muscles movements have been encoded based on Ekman's FACS. Measures inspired by the Action Units of the FACS then allow the Machine Learning system to decode facial behaviour in turn allowing emotional response to various stimuli to be picked up consistently and continuously in real time.

The purpose of their work was to develop a real interacting human-computer system, where algorithms written by humans should be able to capture, mimic and reproduce human perception of facial expressions.

Data collected in the course of Dr Sorci's work show that the perception (i.e. labeling) of a human facial expression by a human observer can be affected by a subjective component, which results in a lack of a unique ground-truth. Underpinning the work of Dr Sorci and his colleagues therefore, has been an attempt to ensure that the system is capable of identifying subtle and difficult to identify emotions in a consistent manner. The system's ability to identify the emotional component of expressions such as the top most photo in figure 1 is a particular strength of the system.

In addition, in the last decade, works on psychophysics and cognitive psychology have shown that face recognition and perception of emotions rely on ‘featural” and ‘configural’ information. Human’s visual perception of a face involves the processing of both local facial measures and their holistic spatial layout. The implication of these findings is that an automatic system, aiming at interpreting faces, needs to extract and make use of these two sources of information as well. The nViso system is designed to meet this need for the dual sources of information. (See figure 2.)

FIGURE 1, MEASURING SUBTLE AND STRONG EMOTION STATES

FIGURE 2, SOFTWARE MAPS 143 POINTS IN THE IMAGE DIRECTLY TO FACIAL MUSCLES

FACIAL IMAGING - IN OPERATION A core point about the work of Dr Sorci, as operationalized by the nViso software system, is that the framework can be adapted to utilize very simple technology, namely remote data-collection using webcam and similar methods of video recording. As the penetration of computers with webcams is increasing rapidly, this opens up the exciting possibility of obtaining “neuroscience” types of results within the context of larger sample, in-home surveys.

Essentially the system runs by streaming/downloading a video (of a TV commercial, promotion, movie trailer and so forth) to a respondent’s computer, where the respondent views the video with their webcam enabled. The webcam records their facial expressions and sends this to a central server for processing. The software analyses this video, providing “emotion scores” for seven emotions on a second-by-second basis. This process is essentially automatic, meaning that in-depth analysis of reaction can be provided very quickly. The process is described in figure 3.

FIGURE 3, NVISO IN OPERATION

An important note for online researchers is that the system can be integrated into a conventional survey, so that standard questions can be asked before and after the video is shown. This allows easy comparison and integration of self-report and passively collected data from the same respondent, opening some exciting possibilities for further research, which we illustrate later in this paper.

Data collection via webcam, whether online in home/work or offline/online in a central location or theatre setting, is not without its challenges, however. These can be categorized in three broad groups:

 Technical, e.g. programming, validation, comparison to 'expert' coding;  Practical, e.g. band width, lighting, eyewear, facial hair;  Data processing, volume, visualization, delivery.

Clearly, all these needed to be addressed and overcome before any analytical approaches were developed.

Technical challenges Figure 4 shows the 'spider' mask superimposed on a subject's face. This is the tool which in identifying and then linking the key muscle points on the face allows the facial expressions to be captured and encoded, thus facilitating the development and application of metrics that allow the processing, aggregation, and interpretation of the facial imaging data collected.

Key to the viable application of the method is its ability to consistently measure the same facial expressions over the full range of physiologies. In other words, to show that a particular set of movements displaying a given emotion or set of emotions will be recorded on one respondent the same as any other displaying the same emotions, despite superficial physical differences.

Considerable effort has been put into ensuring that the measurement of the face is detailed, robust and a valid reflection of the viewer’s underlying emotional reaction. The details of this, validation efforts, and the like are beyond the scope of this paper which is focusing on the practical implications of such methods for online research, but references are appended to this paper and may be obtained from the authors.

FIGURE 4, NVISO FACIAL IMAGING “SPIDER MASK”

Practical challenges While developing a system capable of recognizing and measuring emotions expressed on the face is challenging in itself, doing this in the real life environment of online surveys presents equally challenging practical issues. In reality, most of the images captured will not be in 'ideal' studio like lighting environments, where the face of respondent is perfectly lit and stable on ensure flawless capture. Certainly in the case of in home interviewing the range of environments, light sources, posture, and so forth are heterogeneous.

Computer scientists have been studying automated facial analysis for the past 20 years and this daunting task has become easier. Now the technology exists to overcome the aforementioned practical issues in capturing expressions in natural environments with sufficient precision to allow automated analysis. Computing power and algorithms have advanced enough to make this viable now and into the future. The software has to be sensitive and powerful enough to a capture the hundreds of measurement points involved in tracking the movements of 43 facial muscles in real-time in demanding natural environments.

Furthermore, it has to deal with common features of physical appearance. It is reasonable to request that respondents do not speak, eat, or smoke while watching the stimulus material. These activities distort the muscles whose movements display the respondents' emotional responses. However, the software has to be robust enough to overcome challenges that are everyday such as spectacles/eyewear and facial hair.

Figure 5 demonstrates the range of poses, lighting conditions, and appearances with which the software has to cope. In all these cases, the 'spider' mask was able to catch the emotional responses to a set of TV commercials shown.

In essence then, it is possible to gauge facial reaction to videos with the most basic of equipment, and in a variety of environments.

FIGURE 5, RANGE OF POSES AND LIGHTING CONDITIONS

FACIAL IMAGING – WHY IT MATTERS To illustrate the use of this sort of system in research, and to show what kind of outputs are possible we will draw on examples from case study material from China, New Zealand, and the United Kingdom.

Since the purpose of this paper is to illustrate applications and reasons for using facial imaging in online research, we will draw on a range of examples, rather than showing a complete case study. These examples are all drawn from evaluation of television material.

Example 1 - Heineken 'Walk in Beer Cupboard' – Masses of information, collected passively For this case, a sample of 150 respondents was recruited for interview on line and asked, after appropriate instruction, to view a television commercial. During the viewing part of the interview, the respondents, with their permission, were filmed. The commercial was the acclaimed Heineken 'Walk in Beer Cupboard' commercial.1)

The nViso facial imaging software measured the emotional responses, from facial movements, via webcam, collecting second by second changes in emotional 'expression' for the 150 respondents.

Figure 6 shows both how the emotional intensity overall rises and falls from the neutral baseline over the 30 second period. It not only shows the overall levels, comparable to arousal/activity measurements via a skin sensor (e.g. via galvanistic skin response), EEG, or fMRI but adds markedly more colour by displaying the relative intensity levels of the key emotions.

Thus, we are able to see what is happening with respect to emotional response, as and when it occurred, on a second by second basis. There is generated quantifiable information on “amount” of emotion evoked. This commercial turned out to be highly emotional in contrast to others tested which showed lower peaks. In other words, there is a clear guide to build and decline trends for differing emotions; there is a precise read on “tipping points” where response changes emotional direction.

Furthermore, due to the scalability factor implicit in the nViso software application, we are able to cover much larger samples reasonably cost-effectively. Overall measures of statistical accuracy aside, the key benefit is the ability to look at sub-groups and market segments. Figure 7 shows a simple gender comparison with clear evidence that women lose interest a short way into the ad. (In this case, this may not present a problem depending on the target, as we can see the males' interest captured and sustained.)

FIGURE 6, EMOTIONAL INTENSITY, HEINEKEN COMMERCIAL

FIGURE 7, MALE VS. FEMALE REACTION TO A BEER COMMERCIAL (UK)

Aside from the obvious analysis of what images and scenes are related to what kinds of emotional response, and how long it took for the commercial to build emotional response, three features are of interest for online research:

 While other survey based research systems also attempt to gauge the “shape and intensity” of emotional response, the ability of these (mainly implicit) methods to measure underlying emotional response is subject to much debate. Apart from questions of the accuracy of such measures and the impact of “lagged response” (time between the respondent seeing the stimuli and recording response to it), they usually ask respondents to carry out tasks (moving dials, etc.), require extra questions, sophisticated analysis and/or multiple exposures of the same commercial. The above charts require nothing of respondents beyond watching the video.  A mass of data of differing types is collected, simultaneously, essentially “for free”. Every second of the commercial is providing seven data-points (one for each emotion). While this does raise issues of complexity in dealing with the mass of data obtained, it also opens up the possibility of more sophisticated analytics and driver analysis as clients develop data-bases of such information.  EEG and biometric type tests often require specialized equipment and complex post-study analyses. This raises the cost and militates against larger samples. This in turn can make it difficult to produce sub-sample analyses, of the sort noted above.

From a client perspective then, we see in results like this a simple and cost-effective method to assess key issues such as:

 How long it takes a commercial to build emotion – and therefore whether it can avoid “channel switching” early in the ad.  The exact scene where a sub-group (in this case, females) 'turned off' was identified. If this was important for targeting the issue can be addressed specifically and precisely.  If there is a need to create a 15sec version we could help decide what was most important to retain.

Furthermore, taking a global outlook, it is worth noting that this commercial has been run in several countries in their respective languages. Should international comparisons/re-assurances be needed, the entire test can be run without verbal/written questioning in all key markets to ensure consistency of response/appreciation and so forth. In other words, rather than translating and comparing many different emotional attribute statements, or aligning implicit “metaphor” or similar scales to avoid cultural bias, we can simply show the marketing stimulus and passively collect comparable response.

This has the potential to cut through a considerable amount of effort and debate about cross-cultural norms and scales, and the meaning of terms in multiple languages.

Example 2 – Comparing similar commercials. Fast, comparative feedback from a short, simple survey One of the core issues facing modern researchers is the fact that in the same category it is not uncommon for brands to advertise with very similar benefits, product features and even the look and feel of their advertising. Answering the question of which brand’s advertising creates the most emotional response is difficult.

Facial Imaging collects reaction directly, while viewing is happening, so respondents do not have issues with recall and cannot get “confused” between similar alternative executions as can happen in standard ad-testing procedures.

Hence it is possible to put a series of videos from the same category side by side and assess reaction. Similarly, benchmarking is made easier in terms of comparing test material with direct competition or previous executions. This is potentially much more precise than comparisons with disguised or general “norms”.

The following example is from a study, carried out in conjunction with Saatchi & Saatchi, of New Zealand bank commercials and it illustrates how the system can be used to look across a category at a series of ads. In this case approximately 50 bank customers (n=52-57 depending on the video). Each watched the same seven commercials (six bank ads and one control ad – a Toyota commercial known to score highly on Saatchi & Saatchi’s key “Lovemarks” attributes). As with the Heineken study, respondents were drawn from an online panel (Research Now), were not pre- qualified and simply responded to an invitation to participate in an online study that required them to be filmed.

Figure 8 shows a baseline comparison on a score called “Emotion Lift®” which we have devised as a summary measure for the total emotional content (of positive or negative valence) across the emotions measured. In essence we use it to distinguish emotionally rich from emotionally inert marketing material and it represents the first step in analysis of such stimuli. (We would argue that an ad that is emotionally inert is probably ineffective on any level). It is not our purpose here

to discuss the construction of, or to justify this measure, rather we want simply to use it to show that facial imaging can directly compare multiple ads in terms of emotional response and make clear summary judgments about their content. That there will be argument about precisely the best way to summarize such data, we have no doubt – but this has ever been the case for all measures of advertising effectiveness, and this has at least the advantage of being based on a direct measure of emotional reaction, not self-reported “feelings”.

FIGURE 8, RELATIVE EMOTION LIFT® - SIX NEW ZEALAND BANK ADS

Clearly, such cross category comparisons can be extended to encompass not merely overall assessments of the “amount” of emotion in a commercial, but can compare commercials in terms of specific emotional content. Figure 9 shows the amount of “Happiness evoked” by three of the commercials assessed. It should be noted this is a relative measure of emotion, calculated in terms of each respondent’s “prior emotional state” before watching the commercial. This is, in itself, an important advantage of this technique over self-report or implicit methods of assessing emotion, in that such methods have difficulty in defining a “baseline” emotional state, yet how one feels prior to viewing a video surely impacts how that video is assessed.

FIGURE 9, THREE NEW ZEALAND TELEVISION ADS – SHAPE OF HAPPINESS REACTION

Interestingly, all three of these are long ads (45-60 seconds) seeking, in differing ways, to produce an up-beat, amusing and mostly positive tone. Yet the build and shape of happiness response is quite different for each. Further analysis of

these commercials helped interpret why the build differed, and how particular images, voice-overs and scenes contributed to their ability to build and sustain happiness.

One of the cornerstone analytics in such diagnosis is the ability to drill down and identify periods in the video when emotions change direction and the tone of viewers response changes significantly. We call such periods “Linchpin Moments” as it seems the whole direction of response to the video can turn on reaction to a few vital seconds.

Figure 10 shows how we can use Facial Imaging to drill down and describe the emotional profile of a bank ad and illustrate the changes in reaction. Here we see a few seconds of reaction where happiness and surprise are superseded by sadness, leading to an ad that leaves an overall sad response pattern (which we believe is unlikely to have been the intention of its creators). Further diagnosis of this period revealed almost certainly unforeseen issues with a specific scene (for a very short instance it looks like a bus may collide with a tractor!) and a possible disconnect between the message of the voice-over (which was central to the commercial) and what people were taking from the visuals. A core point in this though, is that this diagnosis is only possible if emotional reaction to a video can be analyzed with precision – we deem it unlikely that the unexpected microsecond of response to a very specific part of a video could have been recalled or described using conventional methods.

FIGURE 10, LINCHPIN MOMENT IN A NEW ZEALAND BANK AD

Example 3 – Potential for integration with self-report data So far in this paper, we have focused on the "special virtues" of facial imaging data, and how it provides unique advantages over conventional question based data. However, it is not our contention that self-report data is not “real” or necessarily inaccurate. When a respondent reports “liking” an ad, or describes it as “impactful”, they are not usually lying: these are statements with meaning. The issue is usually in trying to understand the real level of “liking” or the under-pinning of the reported “impact”. We believe analyses that combine direct measurement of emotional response with self-report measures may actually enhance the value of both.

In this final example, taken from a study undertaken in China with the help of SSI Inc, and Fonterra, we examine reaction from a study of 133 women 18-45 years of age. The ad involved is for Anmum, a brand of milk based nutrient supplement, targeted especially at helping women who have recently delivered babies or are expecting one. It seeks to illustrate how the product benefits aid the development of the growing fetus, in a tone that is positive and emotionally warm. Our results show that, among our sample, the commercial “works” emotionally, delivering a positive response that builds in a sympathetic way with the story the commercial seeks to convey.

More pertinent to this paper, however, is the fact that in evaluating this commercial we were able to compare direct emotional response to standard questionnaire based approaches. Figure 11 shows how “happiness” response varies by self-reported liking for the commercial.

FIGURE 11, RELATIVE HAPPINESS SCORES BY SELF-REPORTED “OVERALL LIKING”

Unsurprisingly, those who like the ad a lot (n=25) tend to show higher happiness scores than those who report themselves neutral or disliking it (n=47). This is especially true near the end of the ad which, given a recall effect, is the part most likely to be remembered in such questions. Interestingly though, the people in the “mid-category” (“Quite Like”) exhibit the lowest scores. It would have been expected that respondents choosing the in-between category on an ordinal scale would also show “in-between results” when emotion was measured directly. We are loath to comment definitively on this result, given the small sample size, until we have compared this with data from other markets. However we are confident this is not a translation effect, and it may be that how we interpret people’s reaction to what seems “obviously” to be ordinal scales will be modified by this kind of research.

FIGURE 12, RELATING IMAGES TO EMOTIONAL RESPONSE

Aside from this speculation though, we can see that “like a lot” respondents actually have a profile that differs considerably from those self-reporting overall negative/neutral reactions. Figure 12 shows an example of a chart where we identify which seconds seem to be really driving the difference between the two categories. What points in the commercial makes one person ‘like it a lot’, while another remains indifferent?

One particular point is important here. Around 18-19 seconds into the ad a stylized image of a fetus in the mother is shown – at this point the “like a lot” group actually shows a huge drop in happiness, while the others remain relatively constant. The happiness of the “like a lots” later rises fast, and it seems as if this drop (possibly it evokes some concern?) is almost necessary to provoking the subsequent rise – it seems it is part of their involvement in the story.

While it would take more research and thought to understand what the client could do to further increase the number who “get the story”, the core point here is that we are seeing signs that by analyzing self-report data in conjunction with emotional response information we can substantially improve the diagnosis offered to clients.

Our view is that this sort of analysis, at the creative development stage of an advertising campaign, can provide focused, impartial, and cost-effective feedback to an agency's creative team allowing them to finesse the commercial and thus increase the likelihood of it being effective in the way intended.

FACIAL IMAGING - DEVELOPMENT POTENTIAL Clearly, based on the above examples, Facial Imaging can offer clients some interesting and unique advantages in terms of marketing outputs. Yet we would argue that the advantages are not entirely about “sexy outputs”. In fact this technology counters some of the key criticisms increasingly being raised about Conventional MR techniques:

 That they are predicated on language/rational models of information gathering, requiring respondents to engage their 'rational' brain.  That standard research is too slow and insensitive when it comes to emotional research, with a reliance on recall and failure to capture fleeting emotional responses in real-time.  The need to improve recording nuances and depth of emotional response.  That debate between various types of scales or methods of measuring emotive response, while fierce, remains controversial and unresolved despite years of analysis of competing systems.  That many evaluation methodologies, even if deemed effective, require too much effort of respondents and are often complex/slow to analyze.

Bypassing the problems of self-reported emotions by relying on passive methods to collect emotional response means we can overcome these challenges and contribute greatly to the development of market research. While we have focused on reaction to video material, in our view, the facial imaging methodology could be adapted and applied to a wide range of marketing research issues, e.g.

 Emotional response to advertising, logos, and related branding  Concept screening and Product/Sensory tests  Product launch prediction/Simulated Test Marketing  Website and online advertising impact  Tracking advertising wear-out  Customer Satisfaction at point-of-service

As noted above, not only does Facial Imaging provide a new kind of leading edge analytics, but it does this in a way that does not require market researchers to abandon other useful and tested analytics. It is perfectly possible to include, in the same survey, both direct measurement of emotional response with 'normal' questions on behaviour, lifestyle and attitudes, for instance. This implies better profiling of emotional reaction and even more excitingly opens up the prospect of very accurate comparison of immediate emotional response with the results of questions on longer term behaviour and motivations. The opportunity to compare the impact of short term stimuli with longer term drivers of equity and loyalty should be greatly enhanced, for instance.

HOW FACIAL IMAGING CAN DELIVER LOGISTICAL AND IMPLEMENTATION ADVANTAGES The benefits noted above are available, in part, because of the method’s inherent logistical and implementation advantages over comparable neuroscience and biometric methodologies.

Scalability One of the main hurdles in augmenting the existing array of market research tools with the benefits of neuroscience applications ('neurometrics') has been essentially logistical. In the main the equipment needed for many applications is expensive, not always highly portable, and often requires specialists or highly trained technicians to operate. Even manual facial coding which has less technology attached can only work when responses are interpreted by dedicated individuals with years of training and experience. Our argument is that facial imaging changes this dynamic as illustrated in figure 13.

FIGURE 13, SCALABILITY OF DIRECT METHODS OF RECORDING EMOTIONAL RESPONSE

Online research recruitment and quality Another benefit of this methodology is on panel and data quality. Put simply, we can see our respondents. This means we can (and do) remove from processing individuals who did not seem to be actually watching the video at all, or who did not follow key instructions. More interestingly, this seems to represent a good method for assessing panel quality – we have seen examples of males sneaking into “female only quotas” (perhaps surprising in that they had to ‘opt in’ to being filmed!), old people pretending to be young or people who do the same interview a couple of times from different locations. Some of this can be automated – for instance it is relatively easy to collect information on whether a person was actually looking at the screen when a video or concept is shown, and this could be extended to gauge attentiveness when answering

questions. Using the data collection capabilities of this system could, therefore, markedly boost efforts to enhance online panel and questionnaire quality.

All this is not, of course, to argue that facial imaging is without logistical challenges. Some of those were noted earlier, but one other core issue is that of recruiting respondents with webcams who are willing to be filmed.

Our experience, though, is that this is not as hard as might be imagined, and that – as shown by the above examples - studies can be undertaken quite easily in a variety of markets. In general any loss of respondents for lack of webcams can be offset by the fact that the questionnaire is potentially a lot shorter (no need for complex questions on emotional response) and the task easier for respondents (“watch some ads”).

This latter advantage is, we would argue, no small one – questionnaire length and complexity is a central part of a research agency’s cost structure, and declining response rates is a key issue. We are finding the idea that “all you have to do is watch” is quite appealing to many panelists when compared to answering batteries of attribute statements or contemplating a complex grid question.

Yet, even accepting the above recruitment challenges, it is critical to understand that these are issues that can be resolved firstly by time, as webcam penetration rapidly increases and spreads to more devices (e.g. iPads, Smartphones) and secondly by training and pre-qualification of panel respondents. We believe that research agencies, community panel providers, and panel companies who see the advantages of this system will rapidly partner to create dedicated sub-panels that can be easily accessed by this technology. This will open up the prospect of speedy, highly cost-effective and mass market assessment of emotional reaction to a wide variety of material. Ultimately, it could transform our industry.

SUMMARY AND IMPLICATIONS FOR MARKET RESEARCH At first glance there is something almost “science fiction like” about facial imaging. The idea that we can remotely measure emotion without asking any questions seems almost unnatural. Yet we would argue that this method in fact takes Market Research back to the most basic of methods by which humans judge emotion – by reading the reaction on someone’s face. If you have a dinner guest, and say to them “I hope you liked dinner?” you are likely to place more faith in the expression on their face than on any polite verbal reassurance they give you – we would argue that the same can, and should, apply to survey response. We have attempted to show that Facial Imaging can be applied to make this most subjective of assessments available in the context of everyday online research.

Importantly, given the current debate surrounding ‘neurometrics’, we believe that the impact of so-called brain science on the online survey need not be so much to displace conventional research but that, properly utilized, methods such as facial imaging have the potential to reinvent and reinvigorate survey methodology. Our work with this method has shown that the potential advantages of automated facial imaging include:

 That Facial Imaging is a sophisticated and robust approach that can be integrated easily into existing survey frameworks,  That it solves some core measurement issues inherent in conventional research approaches,  That by combining Facial Imaging in the same questionnaire as other kinds of questions, there is the potential to markedly improve the quality and marketing usefulness of research outputs.

These advantages, in turn, result not only from the system’s almost magical ability to measure emotion simply by “watching a face’“, but also because, compared to alternatives, this science based approach also has core logistical strengths.

In essence, Facial Imaging is:

 Scalable – so can be used with large or small samples, cost-effectively,  Passive – respondent does not need to do anything except watch, so ‘non-intrusive’,  International – no translation issues, no problems of comparing scales, and  Adaptable – easy to integrate with conventional studies, so useful for many issues.

Properly utilized then, we argue, the system has the potential to reinvigorate and revitalise conventional online research; providing marketers with immediate in-depth emotional response to stimuli, and researchers with a new way to enhance and simplify the task we ask of respondents.

This then, is a genuinely transformational technology.

FOOTNOTE 1. Those interested can find it on http://www.youtube.com/watch?v=wjkUkqTUHNc

REFERENCES Bartlett, P. M., M. Machine analysis of facial expressions, in: K. D. . M. Grgic (Ed.), Face Recognition, Vienna, Austria: I-Tech Education and Publishing,( 2007), Ch. 20, pp. 377–416. Cabeza, R., and T. Kato, Features are also important: Contributions of featural and configural processing to face recognition, Psychological Science 11 (2000) 429–433. Cohn, J. F. Foundations of human computing: facial expression and emotion, in: ICMI ’06: Proceedings of the 8th international conference on Multimodal interfaces, ACM, New York, NY, USA, (2006), pp. 233–238. Darwin, C. The Expression of the Emotions in Man and Animals. J. Murray, London. (1872) Ekman, P. and W. V. Friesen, Facial Action Coding System Investigator’s Guide, Consulting Psycologist Press, Palo Alto, CA, (1978.) Ekman, P., W. Friesen, J. Hager, Facial action coding system, Research Nexus, Network Research Information, Salt Lake City, UT, (2002.) Ekman, P. An argument for basic emotions, Cognition & Emotion 6 (3) (1992) 169–200. Farah, M. J., K. D. Wilson, M. Drain, J. N. Tanaka, What is special" about face perception?, Psychol Rev 105 (3) (1998) 482– 498. Fasel, B. and J. Luettin, Automatic facial expression analysis: A survey, Pattern Recognition 36 (1) (2003) 259–275. Gordon, A. The Face as the Ultimate Survey Tool, Research World Connect June (2011) Keltner, P. and D. Ekman, Facial expression of emotion, in: Handbooks of emotions, M. Lewis and J.M. Havilland, (2000), pp. 236–249. Matsumoto, D. and B Willingham, The thrill of victory and the agony of defeat: Spontaneous expressions of medal winners of the 2004 Athens Olympic Games. Journal of Personality and Social Psychology. (2006) 91(3). 568-581

McCallum, D, Gordon A., and Sorci, M. Say it to my Face - Applying Facial Imaging to Understanding Consumer Emotional Response, Paper presented to AMSRS annual conference, Sydney, September, 2011 Meulders, M., P. D. Boeck, I. V. Mechelen, A. Gelman, Probabilistic feature analysis of facial perception of emotions, Journal Of The Royal Statistical Society Series C 54 (4) (2005) 781–793. Schwaninger, A., J. S. Lobmaier, S. M. Collishaw, Role of featural and configural information in familiar and unfamiliar face recognition, in: BMCV ’02: Proceedings of the Second International Workshop on Biologically Motivated Computer Vision, Springer-Verlag, London, UK, (2002), pp. 643–650. Sorci, M., G. Antonini, J.-P. Thiran, M. Bierlaire, Facial Expressions Evaluation Survey, iTS (2007). Tian, Y.-L., T. Kanade, J. Cohn, Facial expression analysis, in: S. L. . A. Jain (Ed.), Handbook of face recognition, Springer, (2003.)

THE AUTHORS Alastair Gordon is Managing Partner, Gordon & McCallum, New Zealand.

David McCallum is Managing Partner, Gordon & McCallum, Australia.

Matteo Sorci is Co-Founder, nViso SA, Switzerland.

Tim Llewellyn is Co-Founder, nViso SA), Switzerland.

PART 3: SOCIAL MEDIA RESEARCH DIMENSION

THE ROSETTA STONE MEETS FOUCAULT UNDERSTANDING SOCIAL MEDIA VIA DISCOURSE ANALYSIS Ray Poynter

PREFACE When the Rosetta Stone, which dates back to ancient Egypt and the rein of the Ptolemies, was re-discovered in 1799 it provided an invaluable clue in deciphering Egyptian hieroglyphs. The Rosetta Stone contains a decree from Ptolemy V, with the decree repeated in three different scripts (Ancient Egyptian hieroglyphs, demotic script, and Ancient Greek). Understanding the message in one language helped understand its transcription in other languages/scripts. One can see the Rosetta Stone as a model of structuralism (in the sense of Swiss linguist Ferdinand de Saussure), where language was seen as a static system of interconnected units – the modern equivalent being the tourist’s phrasebook.

Michel Foucault was French social theorist and philosopher who in the second half of the 20th century, along with a number of other innovative thinkers (many of them French, such as Levi-Strauss and Barthes) challenged models of thinking. This upheaval in traditional thinking produced an approach that has been labelled by some as post-modern and post-structuralist. Foucault developed a type of discourse analysis that focused not on the words used by individuals, but on language at a macro level, looking at how language was involved in the creation of meaning and power, and how power was involved in the creation of language.

If the Rosetta Stone’s message was that we can find devices/tools (e.g. automated sentiment analysis) that will reveal the meaning communicated by language, then Foucault’s message was that language is fluid, that it creates meaning (rather than simply transmitting meaning), and is both contextual and constitutive. In seeking to understand the discourse of social media this paper will present the case that currently the focus is too much on the Rosetta Stone approach, trying to build lexicographies of meaning, and not paying sufficient attention to the thinking that has been generated by discourse analysis.

INTRODUCTION Discourse Analysis is a relatively modern field, with most of the breakthroughs and developments occurring in the latter half of the 20th century, although some of the seminal work can be traced back to slightly earlier figures such as Mikhail Bakhtin and Ludwig Wittgenstein. Discourse Analysis is not a single method or a single approach to language, rather it is a family of approaches, usually called traditions. These traditions range from Conversation Analysis, a relatively objective assessment of the phenomena utilised in discussions through to activist approaches such as Critical Discourse Theory and socio-historical approaches such as those developed by Foucault and Bakhtin.

This paper starts by providing a quick overview of Discourse Analysis and some of its key ideas before going on to explore how Discourse Analysis can be utilised by researchers wishing to work with the discourse of social media.

Scope When analysing social media we often employ terms like ‘saying’, ‘posting’, or ‘conversation’. However, it is important to note that the term ‘discourse’ is very broad, encompassing media, art, news, fashion, advertising and much more. It is important to note that words like conversation or saying, when used in the context of Discourse Analysis, are typically being used as shorthand for discourse.

A conference paper cannot be a full introduction to Discourse Analysis. This paper concentrates on those features that are common to most Discourse Analysis traditions. This will be followed by a more complete review of a smaller number of specific approaches and the ways in which they can help researchers work with the discourse of social media.

Discourse is more than language The structuralist view of language is that, by and large, language exists and people use it to convey meaning, choosing those units that enable their actions and messages to be understood. However, Discourse Analysts, such as Stuart Hall (1997), make the point that discourse is “not purely a ‘linguistic’ concept”. Hall and others propose that discourse combines both language and practice, that it overrides the traditional split between what we say (language) and what we do (practice).

Two of the key features of discourse are that they are contextual and constitutive. For example my aunt and son might both come out of the same cinema and referring to the same movie say “That film was wicked!” – but they might well mean very different things. When we say words are constitutive we mean they make things, rather than just describing them, for example when you order a meal in a restaurant it causes the meal to be cooked, it does not just describe a process. This knowledge of the contextual and constitutive nature of discourse can allow managers, marketeers, and market researchers to advance beyond a simple intuitive knowledge of ‘what words do’ to a more systematic and powerful utilisation of discourse.

AN INTRODUCTION TO DISCOURSE ANALYSIS There are no agreed definitions of what exactly Discourse Analysis is, indeed some of the traditions of Discourse Analysis are highly critical of the approaches and paradigms of other traditions. However, there are four themes which all the traditions share and which help separate Discourse Analysis from other approaches. These four are:

1. That discourse is constitutive. 2. That discourse is situated and contextual. 3. That discourse is dialogical and contested. 4. That the study of discourse is a proper subject of study, not just a proxy for some underlying phenomenon.

The traditional model of language, for example as advanced by structuralists like Ferdinand de Saussaure, was that it was essentially referential, it was a neutral reporter of events and ideas, a process described by Stephanie Taylor as “transparent or reflective” (Taylor, 2001). As mentioned above when meal is ordered, it is not just describing the meal, it causes the meal to be created. The idea of language being constitutive goes well beyond simple creations, such as jokes/orders/requests, into the creation of ideas, memories, behaviours, and rituals.

The example of the meaning that my aunt and son might be creating by the use of the word ‘wicked’ illustrates the contextual and situated nature of discourse. By contrast structuralists would ascribe the rules of usage as belonging to the language (including dialect and argot) rather than the social.

The Russian philosopher and semiotician Mikhail Bakhtin used the term ‘dialogical’ to express the idea that everything that is said is uttered in response to things that have previously been said. Bakhtin was not just referring to conversations, but to the widest possible interpretation of discourse. Bakhtin’s view places language and discourse firmly within society, and implies that people are both constrained in terms of what can be said and at the same time play a role in the development of how discourses develop. For example, even when we talk to ourselves in our head, or when we write in our diary, we are responding to things that others have said, we are using words that others have said, and we are using meanings that have been passed to us.

The idea that language is contested recognises that discourses are not created and changed by the just the speaker. Listeners have to respond to discourses for changes to happen, phrases have to be repeated for them to become embedded, and meanings need to have a degree of consistency for the discourse to persist. The different Discourse Analysis traditions pay differing degrees of attention to the process of discourse being contested, and to how this process interacts with power structures within society, but they all tend to acknowledge that language is contested.

Discourse Analysis traditions all see discourse as a proper subject for study. Before the rise of Discourse Analysis many disciplines used discourse as source material for their studies, but simply as an input to what they did. For example, psychologists use what people say as a clue to some underlying process, linguists study the structure of language and/or the process of translation, and anthropologists and historians have used discourse as evidence of actions, rituals, and beliefs. Discourse Analysts see talk as action. Discourse Analysis is the study what people ‘do’ when they talk.

Tools and approaches of Discourse Analysis Tools and approaches, for example turn-taking, footing, and language games, are often associated with specific Discourse Analysis traditions, however, it is not unusual for the same tools to be used by some or even all of the other traditions.

Turn-taking is closely associated with Conversation Analysis and refers to the way that people tend to know how to take turns in a conversation, particularly in one-to-one conversations. The clues for turn-taking include short pauses, often just 0.4 seconds long, and a knowledge that lists often consist of three items (at least they do in English speaking countries). The power of turn-taking is most evident when it breaks down, for example if people are combining a face-to-face meeting with others dialling in, or in online conversations, where non sequiturs and out of sequence comments abound.

Footing describes the implied relationship between actors (the people being studied) and the subject of a discussion. In many situations what Discourse Analysts focus on are how footing changes during a discourse. For example, Goffman (1981) uses an example of dialogue captured and reported from a press briefing with US President Richard Nixon to show how the footing of a session can be shifted from the ‘serious’ business of a briefing to the winding up and jokey after chat stage. Nixon achieved the transition by teasing one of reporters, who was female, about the fact she was wearing trousers rather than a dress. Nixon’s actions shifted the discourse from a formal, one-to-many, formal communication to a one-to-one observed, light-hearted (if somewhat sexist) communication.

In the world of social media one of the tasks of moderators is to observe the footing of discussions, looking to shift the footing when necessary. The traditional group dynamics, proposed by Bruce Tuckman (1965), of forming, storming, norming, and performing all represent changes in footing the discourse of the group (along with additions to Tuckman’s original four stages, such as adjourning, transforming, and mourning).

The concept of language games was developed by Ludwig Wittgenstein (1953). Wittgenstein showed that people often communicate utilising understood and recognisable patterns, including patterns of actions. Wittgenstein describes a simple language game between a builder and his mate. The only words in this simple language game are the names of types of stones, such as ‘block’, ‘pillar’, and ‘slab’. The builder shouts one of these words and his mate fetches and passes him the right object. There are no other words, for example terms such as ‘from’, ‘please’, and ‘fetch’ are all omitted, but between this builder and his mate the game functions. Part of Wittgenstein’s point was that language is much more than referential, the builder’s language was constitutive in the very real sense that a wall resulted from the words. In social media similar word games are common, for example the use of favourite and retweet in Twitter, along with RT and #FF (Follow Friday).

John Austin complemented Wittgenstein’s language games with his focus on speech acts. Austin’s speech acts include everyday patterns as greeting, ordering, warning, congratulating, and promising. Putting the language games and speech acts together it is possible to see the Discourse Analysis view of interactions as being based on the utilisation of packages of pre-existing speech, rather than thinking of people as crafting conversations from isolated nouns, verbs, and adjectives.

The traditions of Discourse Analysis The different strands of Discourse Analysis are referred to as traditions. The traditions tend to differ in focus, in epistemology, and in their perceived role of the researcher. However, many, probably most, researchers when tackling a problem would draw on more than one tradition. The following list and comments provide an overview of Discourse Analysis.

The key Discourse Analysis traditions are:

 Conversation Analysis  Discursive Psychology  Foucauldian  Bakhtinian  SocioLinguistics  ‘Critical’ approaches

Conversation Analysis Conversation Analysis focuses on the detail of conversations, seeking to understand what people do when they talk to each other. Conversation Analysis utilises detailed transcripts that record pauses, repairs, and stumbles; it inspects turn-taking, footing, and other devices such as adjacency pairs to identify specific patterns which occur in the conversation. Compared with most other forms of Discourse Analysis, Conversation Analysis sees itself more positivist and less constructionist , i.e. it tends to assume that the observer has a relatively small impact on the analysis and that the findings are ‘evidenced’ by the data.

Conversation Analysis was created by Harvey Sacks in the 1960s and early 1970s. Prior to Sacks it had been assumed that individual conversations were too disorganised and haphazard to be worthy of investigation. Sacks showed that individual conversations could be studied, if they were studied in enough detail.

Sacks did his early work by studying the recordings of people ringing an emergency psychiatric centre. The reason he chose these tapes is that they existed, at a time when tape recording was still quite rare. Because Sacks could get hold of these tapes he was able to study them over and over again to tease out the repeating patterns that indicated that people tended to ‘know’ how to play by (and in some cases break) the ‘rules’ of conversations, rules such as turn-taking. Discursive Psychology moves the focus of the investigation away from the pattern of conversations towards the content of conversations. Discursive Psychology rejects the traditional model of cognitive psychology, that everything that motivates us is inside our heads and instead bring the social and, in particular, the discursive into the picture. Traditional cognitive psychology tends to use the words people say as clues to hidden, unseen processes. By contrast, Discursive Psychology treats what happens during discourse as the topic of its investigation, partly because they believe it is impossible to impute underlying meanings and partly because they believe the locus of many of the phenomena of psychology exist in discourse, i.e. they are shared, rather than existing within the individual’s mind.

Foucauldian Discourse Analysis is based on Michel Foucault’s contribution and takes the focus of research away from individual conversations, and from individuals. Foucauldian analysis looks at discourse at the social and historical level. Foucault takes a post-structuralist approach to discourse, arguing that all meaning is created by discourse, asserting that without discourse a thing does not have meaning, and that a regime of truth exists at any period in time and that the regime defines what can and what cannot be said.

Bakhtinian Analysis is based on the ideas developed by Mikhail Bakhtin in Russia. Despite Bakhtin having been active in the field since the 1920s his ideas did not reach a wide audience until the 1960s. Like Foucault, Bakhtin takes a ‘big picture’ view of discourse, looking at how patterns (which he terms genres) determine what can and can’t be said in a particular context. As mentioned earlier in the paper Bakhtin coined the phrase ‘dialogical’ to emphasise that everything that is said (i.e. any new part of the discourse) is said in response to what has been said before.

Sociolinguistics is a field that is already closer to the analysis of social media discourse than most other branches of Discourse Analysis. Sociolinguistics can include quantitative and positivist approaches, for example Yates’ (2001) investigation of internet interactions. Sociolinguistics is the study of how society shapes the use of language and also the use of language impact and shape society. Sociolinguistics can be used in projects that fall outside of a strict definition of Discourse Analysis, and are closer to ethnography, linguistic anthropology, or ethnomethodology.

Critical Discourse Analysis and Critical Discursive Psychology are both examples of an approach to social research where a project is undertaken with a specific agenda, usually an emancipatory one. Critical social research is sometimes known as action research or critical research and includes projects undertaken with a feminist, anti-racist, or anti-homophobic rationale and perspective. Action research (for example Critical Discourse Analysis) is often criticised as running the risk that the views of the researcher causing the evidence to be misinterpreted. However, the methods and some of the thinking that surround the issue of CDA and CDP might be relevant to the commercial world, where brands and services are looking to maximise the ROI for their shareholders and/or the benefits for their customers.

SOCIAL MEDIA ANALYSIS AND DISCOURSE ANALYSIS Social media analysis, the process of learning about brands, services, and customers by listening to the discourse of the blogosphere, is one of the hottest topics in market research at the moment. Although social media research (also known as blog mining, buzz monitoring, netnography, and blogography) is not typically considered part of Discourse Analysis, since commercial social media analysis tends to take a structuralist approach to comments made, assuming that they directly report views and attitudes.

However, it seems somewhat pedantic to hold that the analysis of the discourse of the web (lower case d and a) does not qualify as Discourse Analysis (upper case D and A). This paper urges that the analysis of the discourse of the social web should be informed by Discourse Analysis and that is the point of the rest of the paper.

Conversation Analysis In many ways Conversation Analysis is one of the most accessible forms of Discourse Analysis, with its focus on understanding what people do when they engage in talk. To date most work with Conversation Analysis has been in the context of spoken conversations, most often between two parties, typically in the context of Western society.

The main lesson from Conversation Analysis for social media research is the value of detailed close inspection, with many of the implications being tactical rather than strategic. One field that is worth detailed examination is that of discussions inside communities, be they MROCs or more accessible communities such as MyStarbucksIdea. Areas of investigation that suggest themselves are:

 The number of individuals who take part in a discussion. Early findings suggest that at any one time the number of people taking part (in the sense of posting as well as reading) is limited, with typical numbers ranging from two to ten.  The relationship between members of a discussion and the moderator or lead poster. Early findings suggest that most members of a community respond to the moderator or lead poster, rather than following discursive threads.  The length of loops/references. Early findings suggest that if a conversation persists over time and generates a large number of posts, most new posts relate to recent comments from others, with earlier comments not being referenced.  The number of repetitions and non sequiturs. The number of references to much earlier comments are rare, but repetitions are quite common, perhaps suggesting that the earlier posts have not been read. A non sequitur is a post which does not appear to follow on from the items above, and these appear to be much more common than in spoken conversation.

A thorough investigation of the rules and processes of discourse in communities will enable brands to maximise the productivity of online communities, operating with the right degree of intervention and moderation.

It is likely that the products of Conversation Analysis applied to communities will differ from those of Facebook posts and differ again from those that relate to Twitter; and it is likely that the picture in 2011 will differ from that in 2015, because expectations and patterns are still being created.

Beyond the analysis of simple phenomena, such as turn taking, reference loops, and non sequiturs, there is ample scope to explore how footing is established in an online discourse, how subjects are constructed, and how people avoid problems and issues, for example resorting to a LOL to avoid an argument – suggesting that a topic should either be treated lightly or not investigated too thoroughly.

Discursive Psychology Discursive Psychology asks a question so bold as to be almost subversive! What if the neuroscientists and cognitive psychologists are looking in the wrong place when they are looking for phenomena such as attitudes and memory? What if these phenomena exist not in the head but in the discourse between people? It is important to note, from a commercial perspective, that it is not necessary to completely accept the argument of Discursive Psychology to realise that this alternative view potentially opens up new insights and opportunities.

To over simplify the argument, for reasons of clarity and brevity, Discursive Psychology is suggesting that traditional psychology uses what people say to guess what is going on in the mind. Discursive Psychology dismisses this as currently impossible and possibly misguided, preferring to categorise and study what people say as phenomena in their own right.

Rosie Campbell has argued (2011) that the personae we project in social media are not the ‘real’ us, but are chosen for specific purposes. Discursive Psychology would make a similar point about all of our discourses, whether online, in print, or face-to-face. The route suggested by Discursive Psychology would be to concentrate on the objects and ideas constructed in the discourse, the roles played, and values evidenced.

One of the key points that Discursive Psychology makes is that a large part of memories are constructed and maintained socially; consider the way that a small child might sit with a grandparent hearing tales of his/her parent’s childhood and looking at family photographs. The social web has given a very tangible form to this process, for example the way that photos are uploaded after a night out and the memories are collectively negotiated and constructed, or the central role that Wikipedia is a taking as a social, collaborative memory bank.

Some brands with enormous numbers of Facebook fans are using these pages to explore this shared psychology. For example when Disney asks “Who is the cutest couple?” or Oreo asks “How would you describe Oreos to somebody who had never tasted one?” they get thousands of replies, mapping parts of the collective psyche.

Foucauldian and Bakhtinian Analysis Although Foucauldian and Bakhtinina Analysis differ in detail and philosophy they can be usefully considered together from a market researcher’s point of view. Both Foucault and Bakhtin look at things at the macro level, not at the level of individual conversations, they both recognise that at a moment in time there are things that are sayable and things that aren’t. Bakhtin introduces the idea of genres, language for a specific context, forum, rule set. Foucault concentrates on genealogies, the historical perspective of the link between power, meaning and language.

In the commercial world a great example of this view of language can be gleaned by the way Starbucks help engender a specific vocabulary of coffee including teaching customers how to pronounce cappuccino and popularising the term

barista. A review of conversations on MyStarbucksIdea clearly illustrates the language that is considered correct for the situation, and which can be used in conjunction with Conversation Analysis to look at how ‘authority’ is established and the footing of discussions changed, and the approach of Discursive Psychologist will help explore the stated truths, the unstated truths, the contradictions in the positions take and the way that these contradictions are negotiated.

SocioLinguistics In many ways sociolinguistics is simply a more formal way of approaching the sort of social media listening that has begun to emerge. Among the sorts of things that social media listening might learn from sociolinguistics are:

1. The use of a comparative corpus, to evaluate whether the corpus collected for a specific brand had systematic differences or similarities to the reference corpus – for example British National Corpus or The Bank of English. 2. Examining the genderlect of a specific discourse, are the conversations more masculine or feminine for a specific campaign and does that change over time. 3. Identifying the creation and transmission of memes. 4. Creating typologies of roles within online discussions.

CONCLUSIONS AND RECOMMENDATIONS Social media research practice is currently running ahead of theory, with dubious findings and conclusions being drawn. The massive discourse of the social web is an incredible resource for brands to explore customers and potential customers in new ways. Unfortunately, at the moment the bulk of social media research seems focused on simply counting instances and of applying a primitive form of sentiment analysis.

This paper has argued that there is a case and there are opportunities to turn to Discourse Analysis to put some intellectual and theoretical muscle in social media research. By utilising tools such as Conversation Analysis and Sociolinguistics brands can obtain a better understanding of the current discourse and a better understanding of how to leverage that discourse for their benefit.

REFERENCES Campbell, R (2011) ‘Who are we meant to be 'spying' on?’, Research-Live, 4 May 2011, Link= http://www.research- live.com/comment/who-are-we-meant-to-be-spying-on?/4005120.article, accessed 17 August 2011 Goffman, E (1981) ‘Footing’, Forms of Talk, Blackwell, Chapter 3 Hall, S (1997) ‘The work of representation’, in S Hall (ed.) Representation: cultural representations and signifying practices, Sage. Taylor, S (2001) ‘Locating and conducting discourse analytic research’, in Wetherall et al (eds) Discourse as Data, Sage. Tuckman, B.W. (1965) ‘Developmental sequence in small groups’, Psychological Bulletin, Volume 63, Number 6, Pages 384-99. Wittgenstein, L (1953) ‘Philosophical Investigations’, Blackwell.

THE AUTHOR Ray Poynter is Executive Vice President and Managing Director, Vision Critical, United Kingdom.

REVIEW OF A DECADE OF NETNOGRAPHY RESEARCH IMPLICATIONS FOR FUTURE SOCIAL MEDIA ANALYSIS Michael Bartl • Gregor Jawecki • Jan Henric Stönner • Dominic Gastes

INTRODUCTION “Open Innovation” refers to the opening of the innovation process towards external stakeholders’ knowledge, creativity and skills (Chesbrough 2003). Nowadays in particular consumers are understood as valuable co-creators of products and services (Prahalad and Ramaswamy 2000; von Hippel 2005). A variety of methods and tools are applied to integrate consumers in different stages of the innovation process, such as idea broadcasting platforms (Lakhani 2006), innovation communities (Adamczyk et al 2010; Sawhney et al 2005), mass customization (Piller 2003) or toolkits (von Hippel and Katz 2002). Besides the emphasis of the active role of consumer, all approaches share that they draw from Social Media to get in touch with knowledgeable consumers from all over the world. Especially the ever growing number of consumer- driven online conversations in forums, blogs, social media platforms or other user-generated content websites (e.g. YouTube, Flickr) creates a valuable source of consumer knowledge and ideas for user-oriented enterprises (Füller et al 2006; Sawhney et al 2005).

One method which is used to integrate consumers’ knowledge into the innovation process is Netnography. Netnography was introduced in the late 1990s (Kozinets 1998) and since then has become a popular method in online research. The word Netnography is a neologism constructed from the words InterNET and ethNOGRAPHY. By “listening in” to natural occurring consumer dialog in online communities, Netnography allows to derive unbiased consumer insights. The researcher deeply immerses into consumers’ explicitly verbalized and implicitly existing attitudes, behaviors, preferences, needs and ideas. Today Netnography is used across industries and for a variety of research questions (see figure 1).

Examples of products such as the NIVEA Invisible for Black & White deodorant (Bilgram et al 2011) show the value of Netnography for new product development. However, they also indicate that the deep insights and innovative ideas resulting from Netnography not only represent great potential but also may impose some challenges to companies. For example, Netnography findings may not always correspond to the company’s knowledge or strategy.

Against this background the objective of this paper is to shed light on how Netnography is currently applied by companies and how the findings are used in the innovation process. A detailed understanding of how consumer insights are utilized and further advanced within companies, as well as the internal processes and structures that are conducive to the adoption of the findings, can provide value in various dimensions. Companies which already conduct Netnography projects can find hints on how to apply the method more efficiently. On the other side, companies which are not yet experienced can use the findings as a guideline of what they have to consider when conducting first pilot projects.

Besides focusing on the current application of Netnography, the paper also sheds light on the method’s development over the past 10 years. To the authors’ knowledge such an overview of the key-milestone does not yet exist. Overall, the timing for an investigation of how Netnography is currently used and its role in the portfolio of research techniques seems perfectly right. Netnography now exists for more than 10 years and has been applied in many companies. Given the growth of Social Media and other virtual communication channels it can be expected that Netnography will even gain in importance in the future. (See figure 1.)

WEB MONITORING AND NETNOGRAPHY Before elaborating the emergence and current application of Netnography in detail, a brief distinction between Web Monitoring and Netnography should be made. As Web Monitoring and Netnography are sometimes incorrectly considered as one and the same a differentiation is important for an interpretation of the findings in the appropriate context and for an understanding of how the insights relate to other research.

FIGURE 1, EXAMPLES OF APPLYING NETNOGRAPHY IN DIFFERENT INDUSTRIES

Web Monitoring and Netnography are both based on web information retrieval. However, the two approaches differ in various dimensions, such as level of automatization and quantity of input data. Web Monitoring is a highly automated process in which software solutions constantly track and analyze online content in search for key words to discover patterns, trends, etc. Due to the automated process Web Monitoring is characterized by high quantitative processing speed and the large amount of data included in the research. In contrast, Netnography is an approach based on manual processing of data. The main advantages of Netnography are the depth of results while the disadvantages are the relatively small data sample and the intensive manual work.

In sum, Netnography and Web Monitoring should be considered as two distinct approaches. As the focus in this paper is exclusively on Netnography, the findings need to be understood in the context of qualitative research and manual data gathering and analysis. The paper does not provide insights related to Web Monitoring or other automated web information retrieval systems.

THE EMERGENCE OF NETNOGRAPHY The term “Netnography” was coined by marketing professor Robert Kozinets, who in 1998 introduced the method as a new way of conducting online market research. Following Kozinets’ suggestion that thousands of consumers engaging in online communities represent a promising, at that time untapped, opportunity for research, in the years thereafter increasing - predominately academic - focus was directed towards the method. Two of the key-publications during this time were Christine Hine’s book “Virtual Ethnography” and Kozinets’ paper “The field behind the screen”. Driven by the understanding that Netnography not only generates insights into consumers’ behavior but that consumers’ thoughts and ideas also provide suggestions for new product development, 2004 marks the year of one of the first officially known applications of Netnography in an innovation-related context. More particular, the sporting goods company Adidas applied Netnography to gather consumer insights and identify business opportunities in the field of basketball footwear (Füller et al 2007).

Starting in 2006 the development of Netnography gained momentum both in the academic as well as the business world. In academia, 10 papers were dedicated to the method within one year – more than twice as much compared to the previous year. Also Netnography was included in the Dictionary of Social Research Methods and companies such as BMW, Gore or Beiersdorf conducted first pilot projects. In 2008 the event Netnography 08 was the first international conference which entirely focused on Netnography and which brought together experts from the academic and business world. By the year 2009 Netnography had its “breakthrough” as a generally accepted tool for online research, illustrated among others by multiple publications (e.g. Bowler 2010; Hück et al 2010; Verhaege et al 2009; Verhaeghe 2010). In 2010, Robert Kozinets introduced his first book entirely focusing on Netnography and the NetnoCamp – the first Netnography barcamp – took place. As of today, several companies have been founded which focus entirely on Netnography, Netnography

publications and reports spread all over the world and almost 100,000 sources can be found with Google search for the term “Netnography”.

In sum, especially in the last four to five years Netnography has rapidly gained attention (see figure 2). It is no longer misunderstood as a methodology that only covers consumers’ chit-chat. Rather Netnography is a generally accepted research methodology and has become a standardized tool in many companies’ portfolio of methods. As more and more companies want to move closer to their consumers, Netnography will likely even become more important in the future.

FIGURE 2, THE EMERGENCE OF NETNOGRAPHY

RESEARCH DESIGN The findings outlined in this paper are based on eight Netnography projects in the field of Fast Moving Consumer Goods (FMCG). FMCG were selected as Netnography has proven a viable tool to support innovation in this highly competitive and dynamic environment. All Netnography projects were conducted for well-known industry players and by an external service provider – a rather typical set up due to the effort and expertise required when conducting Netnography analysis. Each project was mutually exclusive and explored a very own field of research. Further, each Netnography was conducted as a project focusing entirely on online data. This is important to point out, as sometimes the term Netnography is used to refer to projects which also comprise offline in-person observations.

In order to explore the interface of information integration from external sources into the innovation process, the framework of absorptive capacity (Cohen and Levinthal 1990) was used. Furthermore the reconceptualization of the initial thoughts on absorptive capacity by Zahra and George (2002) served as input in the development of an interview questionnaire. The questionnaire aimed to gather profound information from company managers with regard to the utilization of Netnography findings. The guiding questions encompassed the split of potential absorptive capacity and realized absorptive capacity and emphasized the capabilities of assimilation and transformation of information generated with Netnography.

Overall, eleven in-depth interviews were conducted with company managers. All interviewees were involved in at least one Netnography project and were senior managers belonging either to R&D, market research or the marketing department. The Netnography projects commissioned or handled by the interviewees were spread over a period of at least two months and were completed at least four months before the beginning of the research. This time frame allowed the results of the latest Netnography to be adequately handled and processed and enabled the interviewees to give informed responses to the questions asked. All interviews lasted between 45 and 60 minutes. Subsequently, the audio data was transcribed and then subject to qualitative content analysis based on the principles outlined by e.g. Mayring (2002). In the following an overview of the findings is given on an aggregate level. The findings are presented along six key questions, where each question represents a subject area which was repeatedly referred to by the interviewed managers. The main objective in the presentation of the findings is to provide a thorough understanding of why Netnography is applied, how the gathered information is used by managers and which patterns can be identified across projects. The empirical setting differentiates the findings from the conceptual work and case study research mainly conducted so far.

1. Which departments typically conduct Netnography research? Netnography generates deep insights into several aspects, such as consumers’ behavior, language, needs, wishes, perception of existing products and suggestions for improvements. Due to the multitude of findings, it can be expected that Netnography can provide input for various organizational departments. But is the approach for some units better suited than for others?

The Netnographies in the research sample were initiated either by the R&D, market research or marketing department. As will be outlined in more detail in a later part, one of the main intentions in the context of R&D is to gather “fresher” ideas which ultimately contribute to the development of truly innovative products. Also, R&D insights into upcoming trends are of interest. In contrast, when Netnography is applied by the market research department, one objective can be to find out how consumers perceive existing products on the market. Another possibility in the context of market research is to utilize Netnography as a monitoring and optimization tool with focus on specific events in the past. For marketing departments even other aspects are of interest. For instance, insights into consumers’ language when talking about certain products can be used to align the communication strategy of new or already existing products. Another objective can be to evaluate how own and competitors’ marketing efforts are perceived by consumers.

In sum, due to its multidimensional findings Netnography can provide value for various organizational departments. However, the interviews with innovation managers clearly indicated that Netnography yields its full potential when utilized within the framework of product development and innovation. Typically for the R&D team the main objective is to gather new perspectives and inspiration, to spark the internal innovation process and to come up with superior ideas. Therefore the focus is on the creativity, expertise and involvement of consumers that can be met online. Also, for R&D a certain idea is of value independent if it is from one individual or a group of people. In contrast, when Netnography is applied for market research or marketing purposes other questions gain in importance, such as whether the consumers that can be met online are representative for the targeted market, how the authenticity of comments and users can be verified, and whether the relevance of certain findings can be depicted also in quantitative numbers.

Summarized, Netnography yields its full potential in an innovation-related context while for marketing or market research the value is somewhat overshadowed by the method’s restrictions (e.g. non-representativeness, no quantitative validation).

2. What are the objectives when conducting Netnography research? In all observed companies Netnography is used at a very early stage in the innovation process to gather exploratory insights into consumers’ needs, preferences and ideas. The method’s value in the fuzzy front-end is also stimulated by the fact that in the managers’ understanding no other research method can provide comparable insights. Frequently used approaches are either of participative nature (e.g. focus groups) or connected to considerable efforts (e.g. trend research and scenario planning). Also managers stressed that currently used techniques predominately focus on the later stages of the innovation process, such as testing of prototypes. This emphasis is perceived as critical as especially deep insights and good ideas in the early process set the foundation for successful new products.

In their decision to conduct Netnography research, managers are driven either by the desire a) to generate insights for rather new product fields or b) to gather additional knowledge and consumer-inspired ideas for market segments in which the company is already involved. With regard to the first aspect, Netnography focuses on product fields in which the company does not offer products yet but which are considered as promising to extend the firm’s business. In this situation, the company’s desire is to gather a basic knowledge of the market, both in terms of its consumers (e.g. their key needs, preferences, behaviors) as well as commercial players (e.g. key products, brands). The information on consumers and competitive landscape can then be used to evaluate the “fit” between the company and the market, the economic potential and the possibility to develop a product which addresses so-far unsatisfied needs. The method’s application for investigating new markets is also attractive as due to the method’s exploratory nature no prior knowledge is needed. Consequently, without much preliminary work and in a relatively short time span a good orientation in new segments can be attained. While Netnography may not explore all aspects in detail and may leave some questions unanswered, the depth of information is sufficient to decide whether the company should direct further efforts to a certain market. As one manager states:

"And I think that, from my point of view, whenever you enter territories you are not familiar with, which are still somehow uncharted, the core strength of Netnography is its ability to swiftly generate 90% of the knowledge in it."

Besides the intention to gather insights into yet rather unknown segments, Netnography is frequently applied to derive additional knowledge in markets which are already occupied by the company. In this case, Netnography is used to explore consumers’ wants and needs as inspiration for new product development. Naturally the managers in a certain field already have a lot of knowledge on “their” product and market, e.g. from other research activities. Also managers often know

about communities related to “their” product field and even may browse content from time to time. But in their daily routines they typically lack the time to systematically immerse into the online dialog. Against this background, the value of Netnography is that it provides additional consumer insights which can complement other studies and that it gives a condensed snapshot of all existing consumer content on the Internet. Also, the findings can deepen the managers’ current understanding as they stimulate new ways of interpreting already existing data. The following statement by one manager illustrates that it is not necessarily the newness of data that creates value but rather the additional perspective.

"And the way in which the whole thing was processed and aggregated, many aspects in the field of [...] were intensely examined in such a manner that I had not seen before and many points were brought together in a way that enabled me to better understand the topic. Although the technical knowledge was not literally new, it was like a review article which processed the matter very differently."

While the intention to explore new markets or gather additional insights for already known fields are the two main reasons to conduct Netnography research, the interviews also revealed some subordinate benefits that make the method particularly appealing to innovation managers. An overview will be given in the following.

3. What are the advantages of Netnography? All interviewees expressed that one of the main advantages of Netnography is the possibility to obtain uninﬂuenced and freely expressed consumer knowledge. As data gathering is conducted without interference of the researcher in the online dialog the findings are less biased compared to other research approaches. For the managers supervising Netnography research this has the advantage that they are less exposed to critics that the findings may be influenced by their own interests. Consumers’ forthright and genuine opinions also have the “power” to challenge a firm’s assumptions. For instance they can stimulate to reconsider ideas or concepts which were initially considered as “not fitting” to the company.

Another aspect that makes Netnography attractive for innovation managers is the method’s targeted use of the communication channels of today’s consumers. Company representatives are aware that hundreds and thousands of consumers can be met in Social Media and online platforms, such as blogs and forums. In this regard it seems obvious that also new tools are needed that are specifically adapted to the virtual environment and that enable to efficiently access this extensive pool of knowledge. However, Netnography is also not limited to getting in touch with a multitude of consumers at once but in some cases may be used for exactly the opposite – to reach a very small and specific target group. In this case from the online dialog highly expert and involved consumers are identified. They can then be approached to e.g. explore interesting aspects in more detail or to translate insights into product concepts.

The interviews further revealed that Netnography is valued because it is considered a radically new way of conducting online research. While other approaches for online research are based on an active collaboration with consumers (e.g. online focus groups), unlike any other method Netnography derives findings by passively analyzing the openly accessible online dialog. In this sense Netnography is considered as a complement in the researcher’s “toolbox” of methods rather than a substitution of an already known approach.

Last but not least, one benefit of Netnography is that it is still perceived as a new and innovative research method. It stimulates the excitement within the company and compared to other research approaches managers find it easier to create awareness and “buzz” for the findings among their colleagues.

4. What are the method’s limitations and drawbacks? As touched briefly at an earlier point of this paper, one of the limitations of Netnography is that it provides exclusively qualitatively data. While the frequency of how often certain aspects are discussed in communities can serve as a very first quantitative indication of their relevance, still further research is required to quantitatively validate the gathered information. However, in the managers’ understanding the benefits that arise due to the method’s explorative nature outweigh the lack of quantitative data – in particular as a range of techniques are available for a quantitative validation. Similarly, also other aspects which are frequently discussed in literature are not considered as hindering, such as the fact that community members are not representative and can participate in the online dialog anonymously. In the eyes of the interviewees such limitations just induce that in the interpretation of the findings it has to be kept in mind what the method can and cannot deliver.

As outlined briefly at an early point, one aspect that makes Netnography appealing for the innovation process is that it generates truly innovative ideas. However, while creative ideas are per se one of the main ingredients for the development of successful new products, their degree of novelty may even hinder the adoption within the company. Some interviewees reported that not all of the ﬁndings could be utilized, as due to the current company strategy resources could not be directed into projects with more unusual product ideas or with a focus on a speciﬁc target group.

Besides the risk of divergent strategic considerations, the interviews revealed another, much more important aspect that can impede the efficient utilization of Netnography findings. As will be outlined in the following, the adoption of findings is often confronted with mental barriers on an individual level of involved employees.

5. What are the reasons for internal resistance? As Netnography is pure observational research, the results may not always be in-line with the involved employees’ current knowledge, assumptions or interests. Also, due to its explorative nature, Netnography may uncover consumer needs, ideas, terminologies, behaviors, etc. entirely new to the company. As a consequence, interviewees reported that Netnography findings are subject to a sometimes dismissive attitude on the side of involved employees. In this regard the dynamics can be understood as an expression of the so-called Not-Invented-Here Syndrome (Katz and Allen 1982). Not- Invented-Here refers to a company’s strong identification with its own, internal knowledge and reluctance to absorb knowledge stemming from sources outside the company.

With regard to the question whether some dimensions of Netnography findings are assimilated better than others, the interviews suggest a differentiation between two dimensions of knowledge, namely “need information” and “solution information”. While “need information” can be understood as the set of basic requirements, performance expectations and desires consumers can articulate about a product or service, “solution” information comprises consumers’ solution principles and ideas. To provide a better understanding, figure 3 gives an overview of need and solution information for three selected projects.

FIGURE 3, COMPARISON OF NEED AND SOLUTION INFORMATION FOR THREE EXEMPLARY PROJECTS

With regard to the integration of the two dimensions of knowledge the interviews reveal that they are exploited differently by companies (figure 4). Need information is typically regarded as useful and as complementary to the company’s existing knowledge. As need information covers unsatisfied needs, desires or motives it leaves enough interpretational freedom to translate it into possible solutions. In contrast to the open-minded integration of need information, the attitude towards solution information is more ambivalent. Solutions which are shared in the form of first ideas are perceived quite positively, in particular as they still offer some freedom in interpretation and implementation. However, a latent skepticism and threatening perception is often held towards rather elaborated solutions or solutions targeted at core competencies. In this case, company representatives often find it difficult to blind out their own competencies and their prior knowledge. In other words, solution information poses a trigger for potential resistance and denial due to its capability to violate the identity of involved stakeholder departments.

Of all interviewees one third reported that Netnography findings are or at some point have been confronted with a dismissive attitude on the company side. Given this significant risk it is understandable that many of the key success factors when conducting Netnography are related to mechanisms that pursue a better adoption of findings within the company.

FIGURE 4, DIFFERENCES IN THE ASSIMILATION OF NEED AND SOULTION INFORMATION

6. What are the key success factors when conducting Netnography research? To facilitate a smooth transition of Netnography findings a sponsor from within the company is recommendable. The sponsor can address any arising barriers on the side of other employees or teams at an early stage and can ensure that the findings are interpreted in the light of the methods’ peculiarities (e.g. non-representativeness). To facilitate the assimilation in other teams, the role of the sponsor is also to actively involve all internal stakeholders along the process. The early integration of all relevant departments is generally seen as a prerequisite to reduce the risk of barriers due to a conflict of ownership. Ideally other departments are actively integrated in all phases of the development process to let them assess solutions and unfinished concepts and to jointly transform insights into product and service ideas, e.g. in the course of a workshop. In the workshop all participants can bring in their ideas and consequently will feel a higher level of commitment and responsibility towards the output. In sum, Netnography results need more assistance in their transformation and a sponsor can ensure that the right stakeholders are integrated at the right time. In this regard the role of managers changes from supervising single projects to taking the ownership for a certain project along the innovation process.

Another relevant aspect is how ideas are visualized. Although one of the main benefits of visualizations is to achieve a shared understanding, it is advisable to come up with rather “unfinished” designs, such as first scribbles. In contrast to more sophisticated designs, scribbles focus the discussion on solution principles rather than tempting to evaluate particular shapes, colors, etc. Also, scribbles are likely easier to assimilate than user-generated prototypes which convey a rather “finished” appeal.

Companies that intend to conduct Netnography have to consider that considerable expertise and efforts are required for analysis. As typically for every topic hundreds of communities can be found, the challenge is to select the most appropriate ones in the light of the specific research questions. Similarly, the researcher needs some experience to select the most relevant discussions and to derive insights from quantitative data analysis. In the light of these requirements, several interviewees stated that the collaboration with an external service provider can be beneficial. Besides the advantage that the company managers can concentrate on their daily business, working with an external partner also limits the risk that the insights’ subjectivity is called into question. Interviewees also shared some recommendations with regard to the collaboration with an external service provider, for example the importance of the kick-off briefing. Before the start of a project it is required that the company and the agency define the specific research interests. In this process the balance has to be found between deciding which fields are of interest and at the same time maintaining the method’s explorative character. In this regard interviewees also stated that close reporting on a regular basis with constant research focus adjustments can yield positive results.

CONCLUSION Since its introduction to the research world in the year 1998 Netnography has been constantly improved so that nowadays it can be seen as a generally accepted research approach. Various examples show that Netnography is not limited to a certain industry, product category or company size. Due to its multidimensional results Netnography can provide value for various organizational departments such as R&D, marketing or market research. In particular in the context of innovation Netnography yields its full potential as it allows accessing a new source of knowledge and ideas. The insights and ideas derived from communities are considered as more consumer-oriented and “fresher”.

As many companies have successfully conducted first pilot projects, for many firms the focus nowadays shifts to getting a feeling for which of their research questions Netnography can provide value and in which situations other approaches are favorable. Also increasing attention is on the question of how the efficiency when conducting Netnography research can be increased. Overall, methodological expertise is not only a prerequisite for the repeated application of a certain approach but also for the adoption of a long term Open Innovation orientation. In the Global Co-Creation Study (Capgemini 2010),

managers referred to missing knowledge in the application of different methods as one of the biggest barriers in the integration of the Open Innovation paradigm in companies.

Besides methodological expertise Netnography also requires organizational changes. The main challenge when conducting Netnography research is to facilitate an efficient adoption of the findings within the company. In particular solution information is prone to internal resistance. While supporting measures can be taken both by management (e.g. through training) as well as the project sponsor (e.g. through early integration of other departments), the change of mind starts on an individual level. External ideas need to be considered as additional inspiration and food for thought rather than as competition to one’s own skills and knowledge.

The study at hand focuses entirely on projects in the FMCG industry and thus can provide no generalizable insights. For future research it would be interesting to explore whether similar opinions and processes can be found when conducting Netnography in other fields, such as B2B.

Implications for future Social Media analysis In light of the growth of web and social media platforms that provide consumers from all over the world a place to interact, Netnography will likely even gain in importance in the future. Besides increasing the efficiency of Netnographic research the focus will shift to the question if and how new outlets such as Twitter, YouTube, Flickr and Facebook can be included in analysis.

Increasing Social Media research implies that companies should establish the structures and processes that are conducive to the assimilation of consumer insights. Interdisciplinary teams or even the implementation of new departments that focus entirely on consumer communities can facilitate the efficient absorption and transfer of knowledge. Also on the side of innovation managers new skills and capabilities will be needed, such as affinity towards the virtual world or a feeling for when and how to integrate other teams or departments into the innovation process.

In future more and more interest will be on the question of how Netnography can use synergies with other research techniques. In this regard Web Monitoring is of particular importance. While Netnography and Web Monitoring are sometimes understood as competing, both approaches should be considered as complementing each other. The main advantage of Netnography is the deepness of results while one of the disadvantages is the intensive manual work and consequently the required pre-selection of content. The main advantages of Web Monitoring are automatization, high quantitative processing speed and the high amount of data that can be included. On the other side the richness of results is rather limited. Obviously, the advantages of Web Monitoring are at the same time the disadvantages of Netnography. This raises the question of how quantitative and automated Netnography can be without jeopardizing the quality of results. What is the equilibrium of qualitative and quantitative research for Netnography? In other words, how can Web Monitoring solutions successfully be integrated into Netnography and vice versa? In the interviews the innovation managers clearly expressed that they do not want to abandon the qualitative nature of Netnography but at the same time also indicated that some kind of quantification can even increase the method’s value.

Overall, Netnography is just one method which is based on the idea that better products can be developed in a close collaboration between companies and consumers. In the future more and more projects will pursue co-creation between companies and consumers. In this regard, increasing attention will be paid to the people behind the online dialog. For example, the most creative consumers can be identified via Netnography and can then be integrated into further innovation activities, such as Lead User workshops.

LITERATURE Adamczyk, S., Bullinger, A.C., Möslein, K.M. (2010), “Call for Attention – Attracting and Activating Innovators”, Paper presented at R&D Management Conference 2010, Manchester. Bartl, M., Hück, S., Ruppert, S. (2009), “Netnography Research: Community Insights in the Cosmetics Industry“, Conference Proceedings ESOMAR Consumer Insights 2009, Dubai. Bilgram, V., Bartl M., Biel S. (2011), “Getting Closer to the Consumer: How Nivea Co-Creates New Products”, Marketing Review St. Gallen, (1), 34-40. Bowler, G. M. (2010), “Netnography: A Method Specifically Designed to Study Cultures and Communities Online“, The Qualitative Report, Volume 15, Number 5, September, 2010 1270-1275. Capgemini Consulting (2010): Co-Creation Beyond the Hype: Results of the Global Co-Creation Survey, Utrecht, 2010. Chesbrough, H. (2003), “The Era of Open Innovation”, MIT Sloan Management Review, 44(3), 35-41. Cohen, W. M., Levinthal. D. A. (1990) “Absorptive capacity: A New Perspective on Learning and Innovation”, Administrative Science Quarterly, 35(1).

Füller, J., Bartl, M., Ernst, H., Mühlbacher, H. (2006) “Community Based Innovation: How to Integrate Members of Virtual Communities into New Product Development“, Electronic Commerce Research, 6(2), 57-73. Füller, J., Jawecki, G., Mühlbacher, H. (2007), ”Innovation Creation by Online Basketball Communities”, Journal of Business Research, 60 (1), 60-71. Hück, S., Jonas, J., Grünhagen, A., Lichter, C. (2010), “Listening to Social Media from a B2B2C Perspective“. How to Strengthen the Competitive Role as a “Preferred Supplier” with Netnography”, Conference Proceedings ESOMAR Online Research 2010. Katz, R., Allen, T.J. (1982), “Investigating the Not Invented Here (NIH) Syndrome: A Look at the Performance, Tenure, and Communication Patterns of 50 R&D Project Groups”, R&D Management, 12, 7-19. Kozinets, R.V. (1998), “On Netnography: Initial Reflections on Consumer Research Investigations of Cyberculture”, Advances in Consumer Research, 25, 366-371. Mayring P. (2008), “Qualitative Inhaltsanalyse: Grundlagen und Techniken“, Beltz Pädagogik, Weinheim. Piller, F. (2003), “Mass Customization”, Wiesbaden, Gabler. Prahalad, C., Ramaswamy, V. (2000), “Co-opting Customer Competence”, Harvard Business Review, 78(1), 79-91. Sawhney, M., Verona, G., Prandelli, E. (2005), “Collaborating to Create: The Internet as a Platform for Customer Engagement in Product Innovation”, Journal of Interactive Marketing, 19(4), 4-17. von Hippel, E. (2005), “Democratizing Innovation”, Cambridge. von Hippel, E., Katz, R. (2002), “Shifting Innovation to Users via Toolkits”, Management Science, 48(7), 821-833. Verhaeghe, A. (2010), “And they lived happily ever after... Analysing user generated content on social media to increase the elderly's quality-of-life”, ESOMAR Global Healthcare Conference 2010. Verhaeghe, A., Schillewaert, N., Van den Berge, E. (2009), “Getting answers without asking question. The evaluation of a TV programme based on social media”, Proceedings ESOMAR Online Research Conference, 2009. Zahra S.A. George. G. (2002), “Absorptive capacity: A review, reconceptualization, and extension”, Academy of Management Review, 27(2): 185–203.

THE AUTHORS Michael Bartl is CEO, HYVE, Germany.

Gregor Jawecki is Director Innovation Research, HYVE, Germany.

Jan Henric Stönner, Karlsruher Institute of Technology, Germany.

Dominic Gastes, Karlsruher Institute of Technology, Germany.

TELL ME WHAT YOU WANT, WHAT YOU REALLY REALLY WANT CREATING DESIRED RESULTS FROM SOCIAL MEDIA RESEARCH Annie Pettit

INTRODUCTION Social media data is one of the most easily accessible sources of data for both researchers and non-researchers alike. You simply need to turn on your computer, make your way to Google, and search for the name of the product or service in which you are interested. In mere seconds, thousands or millions of consumer comments, messages, opinions, status updates, and tweets from places like Facebook, Twitter, YouTube, Flickr, Blogger and more will appear on your computer screen. It’s far quicker and easier than commissioning an online survey, mobile survey, focus group, or any other form of established research process.

As simple as it sounds to gather data, however, it isn’t that simple to gather quality data. Our collective experience with traditional research methods has taught us that easy and quick do not necessarily generate quality. Poor sampling and weighting techniques, leading survey questions, forgetting to include a Don’t Know or None of the Above, failing to reverse code grids, overly long surveys, and a variety of other survey issues can unite to create a survey project that fails in its ability to properly answer a research question. In the field of market research, basing research on data that is less than high quality can be the difference between political predictions that are accurate or miserably incorrect, or the difference between a product that fails or succeeds.

In this paper, we will demonstrate how the same problem can occur with social media research. We will show how easy it is to draw false conclusions about products simply because care isn’t taken to build a quality social media data dataset. We will showcase the various places where errors or omissions can take place and how severe the consequences can be. We’ll take a journey through a comedy of errors, a case study of what can happen when multiple errors are made in many different places. As a consequence, we’ll attempt to prove that people like dreaded bank fees more than they like yummy, tasty cookies.

DATA COLLECTION The first step of social media research involves identifying and collecting the appropriate set of data. In this case, we are interested in comparing the sentiment towards bank fees and cookies. The data will be gathered from a variety of websites across the internet including blogs, microblogs, video sites, photo sites, forums, news sites, and much more. In order to demonstrate both the correct and incorrect way to gather data, we will show results not for two sets of data but for four. For each dataset, we’ll collect a sample of at least 250,000 records.

1. Cookies (Correct) – This set of data will include verbatims from the social media space as long as they mention cookies that are intended for eating. As such, each piece of data will specifically mention cookies, but additional checks will be added to the data collection system to ensure that each piece of data also refers to things such as chocolate, oatmeal, nuts, sugar, and other baking and eating references. 2. Cookies (Incorrect) – This set of data will gather verbatims that mention the word cookie, but it will fail to recognize that the word cookie can be used in a non-eating way. As such, it will include a wide range of data that specifically mentions chocolate chip cookies or oatmeal cookies but it will accidentally include references to computer cookies, an essential feature of many websites and something with which users have a love/hate relationship. This dataset, therefore, will include mentions of computer cookies, http cookies, server cookies, persistent cookies, internet cookies. Double meanings for words, as with cookies, are not unusual but the meanings cannot be discovered unless specific attempts are made to search them out. For example, everyone knows that the NFL refers to the National Football League, but the dataset would actually include many references to the National Forensic League, Newfoundland, Nerve Fibre Layer, No-Fire Line, and many more completely wrong pieces of data. 3. Bank fees (Correct) – Our third dataset will focus on bank fees such as ATM fees, chequing fees, account fees, and mortgage fees.

4. Bank fees (Incorrect) – Our final dataset will also focus on bank fees. We’ll learn more about the errors associated with this dataset in later stages.

SPAM CLEANING Spam is a massive problem with internet data and in particular, with blogs and forums. To deal with this problem, most blog authors and owners have reverted to moderating comments, or not showing the comments until a human being has verified that it isn’t spam. And, most forum owners require users to create verified usernames and passwords before they are permitted to post comments. Spam comments exist in hopes that people reading them will click on the links because they suddenly realize that they are in fact interested in buying the products mentioned. But, with experience in the social media space, we know that these pieces of data are just one of millions of computer generated messages left indiscriminately everywhere on the internet.

True spam should be immediately deleted without any regard for the content. But, if it is unknowingly left in a dataset, automated processes will assign sentiment scores to it as if it was regular pieces of data. Since these spam messages can exist in the thousands in large datasets, the large volume means they can easily shift a brand score. Spam referencing negative elements such as smoking or ailments will bring brand sentiment down while spam referencing positive elements such as low fees and free will bring sentiment up. Neutral spam will pull a negative brand up and a positive brand down. Spam generally falls into a number of categories, the most popular being pharmaceutical, financial, pornographic, and generic spam. Real examples demonstrating the complete lack of grammatical and contextual coherency associated with spam are illustrated in table 1.

TABLE 1 Type of Spam Example Financial cash advance free and low fee payday loan cash advance capital boulevard guarantee faxless bad credit payday loans, cash payday advance .... fax! definition cash Generic for sale Free porn video sarah jay autumn moon flower sparxxx black Dell zino hd Soft cookie recipe Identity theft Honda diesel generator Special hazards Concealed carry handbag Hid retrofit switched ground How to copy a cd with windows 7 Celebrity Pharmaceutical hgh phentermine quit smoking phentermine no processing fees order phentermine by phone online consultation different types of phentermine pills purchase phentermine .... Pharmaceutical fees brand name nimetazepam no rx order nimetazepam no visa online without prescription text on this WikiDoc page by clicking on .... percocet and zopiclone Without Prescription Pornographic classic cookie monster eating cookies youtube free girl porn movies 3 guys 1 hammer link latin girs get fucked aluminum wicker .... ://phpbb2.altric.com/viewtopic.php?p=2821#2821]better off dead camaro pictures[/url] [url=http://www.comboadictos.net/forum/showthread.php Pornographic Eye surgery in virginia Invention lesson plans inventions by kids Lexapro female sex drive Davis varsity theater plan room Cookie .... plans Other ofac sanctions programs Comments on horoscope other random shit Herbal treatment for prematureejaculation Wellbutrin appetite suppressant Herbal remedies cure (Side note: Because the system used to collect data for this research project has a three stage automatic spam identification system, these types of spam records weren’t even permitted into the raw dataset. As such, this paper cannot show the full erroneous effects of spam.)

RESEARCH OBJECTIVES CLEANING When survey research is being planned, a major portion of the work includes deciding upon the research objectives and designing a survey which will meet those specific needs. The same applies to social media research. Once research objectives have been designed, they will guide the data collection set-up and the type of data that needs to be collected. Data that does not specifically suit the purposes of the research objectives are irrelevant and must be removed from the dataset. As with spam, large quantities of irrelevant data can erroneously inflate or deflate sentiment scores.

Table 2 provides examples of the various types of irrelevant data that might be observed in a normal dataset.

TABLE 2 Type of Irrelevant Data Effect on Verbatim Sentiment Programming Errors Increase areas St. Jude Dream Home, which raises money for children This data mentions cookies, being treated for life-threatening diseases at St. Jude Childrens but it is accidental Research .... -this-happen]').click(function() }); }); programming codes. /features|100737149 Comments (0) Submit a commentloggedIn = readCookie('ccp-user'); if (loggedIn) } else }Click "Report Abuse .... Coupon Listings Increase Huggies Jumbo Packs-$2 MFQ HERE-$2 +UP Reward .... This data is a list of coupons Popcorn-$1 +UP Reward= $0.99$2.99 Nabisco or Pepperidge that a consumer has used, Farm Cookies, Select Varieties-$1 +UP Reward= $1.992/$7 but clients may not be Kotex .... /2 MFQ HERE= $1.49 each wyb 2$2.99 Pepperidge interested in it. Farm Cookies-$1 MFQ HERE-$1 +UP Reward= $0.99$29.99 Viral Games Increase We should talk More [] Cheerios .... Dont even Inbox me ! [] Rice This data is a social media Crispy .... I Love You [] Cookie Crisp .... Inbox Me [] Reese's Puff game that a client may not be ..... Ummm Wassup with me n You ? interested in. Secondary Product Usage Decrease lot like cookie dough ice cream, especially since there are little This data refers to ice cream chocolate chips in it..... water it tastes like sweet water, but with flavoured as cookies, not 2 scoops in 16oz. of milk it tastes like one of those hershey cookies cookies 'n creme bars.

Irrelevant data do not magically raise their hands to identify themselves. Human eyes must review random samples of the data and search out pieces that did not seem to fit. There may in fact be more pieces, but unless someone sees them, they will remain in the dataset and they will affect the overall scores.

Table 3 shows a sample of irrelevant data points and the sentiment associated with each one for the Cookie datasets. First of all, the Clean Cookie dataset has had all spam and irrelevant datapoints removed. As such, there is no spam to report on. On the other hand, the Dirty Cookie dataset has a variety of irrelevant datapoints.

TABLE 3 Dataset Data extract Average Sentiment Sample (1=negative, 5=positive) Size Cookies Perfect dataset, No spam or irrelevant data 4.07 168000 (Clean)

Cookies TOTAL (Warts and all) 3.84 189000 (Dirty) Your browser must be set to accept cookies 4.35 5025 How to clear cookies 4.32 4159 8. Saturday or Sunday? 9. Cupcakes or Cookies? 10. 4.25 311 Chocolate or Vanilla? Recipe 4.25 6969 Cookies and cream 4.17 2461 Allow cookies 4.14 250 Buddy the beagle 4.07 359 Enable cookies 3.88 1515 Cookie sheet 3.86 4574 Cookie cutter 3.69 5566 Save cookies 3.27 53 Delete your browser's stored cookie 3.26 8380 Cookies do not contain any personal information 2.72 678

For our comedy of errors, let’s consider that our lazy researcher saw a few of the mentions of browser cookies. But, they only happened upon a few of them, the ones wherein people used the phrases “your browser must be set to accept cookies” and “how to clear cookies.” There were about 9,000 occurrences of these exact phrases within a variety of longer messages, and since their sentiment was higher than the Clean Cookies sentiment, deleting them from the dataset decreased the overall sentiment of the Dirty Cookies dataset. In addition, our lazy researcher found 311 positive records that reflected a viral game (Saturday or Sunday). Deleting those records also lowered the sentiment of the Dirty Cookies dataset.

Unfortunately, our lazy researcher failed to notice a few additional types of irrelevant data. For instance, there were at least 11,000 additional negative mentions of browser cookies (e.g., “Delete your browser”, “Personal information”) that unknowingly remained in the dataset. Leaving these irrelevant records in served to keep the sentiment of the Dirty cookie dataset lower.

The same processes were then applied to the Fees datasets. Of course, the Clean Fees dataset had neither spam nor irrelevant datapoints. And as before, irrelevant data did not magically raise their hands but rather were specifically sought out by human eyes. Table 4 shows a number of examples of irrelevant data for the Fees dataset.

TABLE 4 Dataset Data Extract Average Sentiment Sample (1=negative, 5=positive) Size

Fees (Clean) Perfect dataset, No spam or irrelevant data 3.52 316000

Fees (Dirty) TOTAL (Warts and all) 3.60 217000 malware 1.63 36 disclosures, prohibit unreasonable fees and the 2.12 66 imposition of mandatory property damage or loss insurance Scam/scammed/scamming 2.14 2675 Debt crisis 2.65 84 Seller fees 2.92 165 EBay fees 3.01 707 No transaction fee (NTF) funds are subject to the terms 3.15 925 Condo fee 3.24 1317 Jetabroad's flight results - no extra charges or credit 3.44 355 card fees The insurance fee is based on your invoice total 3.80 2157 Bankrate.com, Interest.com, Mortgage-calc.com, , 3.86 1382 Savingforcollege.com, Fee Disclosure, InsureMe CreditCardGuide.com, Bankaholic, CreditCards.com .... than 300 financial products, including mortgages, credit cards, new and used auto loans, money market accounts and CDs, checking and ATM fees

In this case, our lazy researcher managed to find some records related to malware, scams, property damage, the debt crisis, eBay fees, and seller fees which had lower sentiment scores and which were deemed not relevant to the objectives of the research. However, our researcher did not find a set of 1,382 duplicate records that listed out banking websites and these remained erroneously in the dataset, increasing sentiment scores.

Because our researcher failed to apply the objectives of the research to the data collection and failed to find all pieces of irrelevant data for both the Fees and Cookies datasets, each datasets contained portions of irrelevant data which affected the results. Specifically, the Cookies Dataset became less positive than it should be and the Fees Dataset became more positive than it should be.

SAMPLING In survey research, an important part of the planning process is ensuring that we reach a representative sample of people. We need to be careful not to miss major segments or our final results will fail to reflect their opinions. The end result would be sample results that don’t generalize back to overall population.

The same holds true in social media research. It’s very convenient for researchers to focus their efforts on a single source like Twitter, but we know that single website sources do not yield research results which generalize beyond that website. One could attempt to collect data from the entire internet space, but very few companies have the resources to duplicate the work of Google. And even Google cannot store the entire contents of the internet (e.g., they can’t see silo websites, or unlinked communities). Consequently, no matter how much effort is putting into collecting data from the full range of websites, some important ones are guaranteed to be missing.

Further, as part of considering research objectives, we need to ensure that our data sources are appropriate for the research. For instance, the researcher may not be interested in off the cuff remarks from microblogs but rather in well- thought out opinions from blogs. Or, maybe they want to focus on news sites in hopes of gathering opinions from more educated, higher income people. Either way, decisions must be made about which websites are appropriate to meet the objectives of the research.

In table 5, we present the sentiment scores associated with data from a number of different websites. First of all, it is clear that sentiment is dependent upon the website source. Data from Facebook is more positive (4.1 out of 5) whereas data from YouTube is less positive (3.2 out of 5).

TABLE 5 Dataset Source Average Sentiment Sample Size Cookies (Clean) Keep data from every website, except sites that are 4.07 168000 pure spam

Cookies (Dirty) http://www.bellinghamherald.com 4.32 4019 http://www.facebook.com/people 4.13 29651 http://gaiaonline.com 3.68 6086 http://profiles.google.com 4.01 1156 http://www.facebook.com/pages 4.43 1332 http://forums.theknot.com 4.01 348 http://www.runnersworld.com 3.82 199 http://www.youtube.com 3.21 1963 http://community.thenest.com 3.74 542 http://www.cpubco.com 2.47 198

GENERAL SENTIMENT ANALYSIS Once again, we begin our discussion in the survey space. Researchers commonly make decisions about how many points should be reflected in a grid question scale, whether 5, 7, or 10 points. Further, they make decisions about whether there will be positively and/or negatively worded items in grid questions, often to the fault of failing to provide negatively worded items. These decisions can have serious implications for how survey responders interpret and respond to the questions.

Of course, the same theory applies to social media research. Though social media research does not use grid questions, it does use sentiment analysis to assign opinions into Likert style formatting, including positive, neutral, and negative groups. Of course, inaccuracies with assigning verbatims to the appropriate point on the scale will result in work that is less valid than desired.

Because language changes so quickly, sentiment analysis is a constantly evolving process that must be kept up to date. Acronyms come and go, and it wasn’t too long ago that FTW, WTF, LOL, FTS, and SMH, had no real meaning. They may not be widely known among non-internet and non-chatting populations but they are in fact widely used words, each with very clear sentiment, and cannot and should not be ignored.

In addition, in order to avoid negative perceptions, some people choose to use alternate spellings for profanity such as F***, fvck, fugg, fawk, or feck, all of which are extremely negative. We know these words are intended to be profanity and so they should be scored as such even though they are spelled ‘incorrectly.’

For our purposes, our lazy researcher failed to learn the most up to date slang and profanity and failed to account for positive slang for cookies and negative slang for Fees (see table 7).

TABLE 7 Sentiment phrases Cookies (Clean) Score all sentiment properly Cookies (Dirty) Fail to score FTW LOL  :] [: =) (= Fees (Clean) Score all sentiment properly Fees (Dirty) Fail to score FTL FTS SMH  :[ ]: =( )= Fail to score an assortment of inappropriate to print on paper slang swear words

BRAND RELEVANT SENTIMENT ANALYSIS Sentiment analysis, like every other stage of the social media research process, is tricky and it can be misled in various ways. For example, the name of the laundry detergent “Cheer” has a positive message implicit in the brand name while the “Texas Chainsaw Massacre” movie has a very negative message implicit in its name. Other brands have issues simply because of how people talk about the product or service. For instance, descriptions of movies, video games, and pharmaceutical products include negative words even though the negative words aren’t comments on the brand. Someone who loved the movie Beloved might say, “I just watched the movie Beloved about a horrible rape and murder. What a fabulous movie!” Or, someone might say, “I get horrible cramps and headaches but Tylenol always fixes me up.”

Part of the sentiment validation process entails ensuring that brand names, taglines, and product descriptions don’t inadvertently affect the overall sentiment of a brand. Unfortunately, this is a yet another step for the researcher, one that requires in-depth knowledge of the brand and a massive attention to detail. But this is essential for the validity of the work.

Table 8 identifies some example sentiment words and phrases that are product descriptions as opposed to product sentiment. For instance, “I hate rainbow sprinkles on cookies” shouldn’t generate a positive mention for rainbows simply because people like rainbows in the sky. And, “The bank handles a lot of bad credit clients” shouldn’t generate negative sentiment because they are able to help people with bad credit. Our researcher simply didn’t have time to look for product descriptions in the datasets and accidentally ended up scoring the sentiment for the following words.

TABLE 8 Dataset Sentiment phrases Cookies (Clean) Score all sentiment properly

Cookies (Dirty) Organic Rainbow Sweet Warm

Fees (Clean) Score all sentiment properly

Fees (Dirty) Low fee Refund

FINAL COMPARISON Our journey has taken us through haphazard data collection, insufficient cleaning of spam, inefficient sampling, and poor quality sentiment scoring. Given what we know about the normal curve, it is highly unlikely that any individual project would suffer from all of them. But, the normal curve does have extremes and this was the unfortunate project that suffered.

To put the final conclusions into a format that will allow everyone to see the effects of the treatments, the figures on the following pages show Top 2 Box scores for three sets of data including:

1. Clean – This data was treated with kid gloves. It was gathered correctly, cleaned of spam, sampled correctly, and fine- tuned for category specific sentiment scoring. 2. Dirty – This data was manipulated as if an ignorant, careless, or time-pressured researcher missed important components of the research process. 3. Raw – This data was not manipulated. It was pulled into the social media research system and not touched by a researcher. It contains all the irrelevant data, all the website sources, and the sentiment was not fine-tuned in any way.

Figure 1 shows three trend lines for cookies. First, it is important to notice that the three lines are very different. Where the Clean Cookie dataset hovers around 48% positive, the Raw Cookie dataset hovers around 42% positive and the Dirty Cookie dataset hovers around 38% positive. This difference of 10% is massive in the market research space and would have serious implications for a brand. It is clearly of absolute importance to identify and remove spam and irrelevant data, ensure that sentiment is scored correctly, and ensure that the right sources of data are identified. Even if no efforts are made to manipulate data, clean data does not generate the same results as raw data.

FIGURE 1, TOP 2 BOX SCORES FOR THREE “EQUIVALENT” COOKIE DATASETS

Second, it is clear that we were easily able to influence sentiment by carefully selecting which portions of the data we wished to clean, bothered to clean, realized we should clean, or were able to clean. Because social media datasets are massive, often numbering in the hundreds of thousands or millions of records, it is very easy to miss a portion of data that needs to be cleaned. And, by failing to account for new slang, we were able to further manipulate the sentiment scores to our advantage.

The same trend was also observed for the fees datasets. Where the Clean Fees dataset generated Top 2 Box scores around 28%, the Raw Fees dataset produced scores of about 12%, and the Dirty Fees dataset generated scores of about 31%. Data cannot simply be collected and measured. Time and effort must be taken to carefully evaluate and fine-tune each process to ensure the most valid dataset is produced. (See figure 2.)

FIGURE 2, TOP 2 BOX SCORES FOR THREE “EQUIVALENT” FEES DATASETS

And now, let’s first consider reality, a direct comparison of fees and cookies using high quality datasets, fine-tuned for the correct data collection, the correct sampling processes, and the correct sentiment system. There is clearly a significant discrepancy in opinions towards fees and cookies. About 48% of verbatims regarding cookies were positive compared to just 28% of verbatims regarding fees. One could conclude that opinions towards cookies are 70% more positive than opinions towards fees. (See figure 3.)

FIGURE 3, TOP 2 BOX SCORES FOR HIGH QUALITY FEES AND COOKIE DATASETS

And now, for the final result. Can a less than stellar treatment result in the conclusion that people like fees more than cookies? With average Top 2 Box scores of 38% for Cookies and 31% for Fees, we didn’t succeed completely. But, unlike our previous comparison showing a difference of 70%, in this case, cookies generate opinions which are only 23% more positive than opinions towards Fees. Cookies are now in stiff competition with fees to retain the most positive position. (See figure 4.)

FIGURE 4, TOP 2 BOX SCORES FOR POOR QUALITY FEES AND COOKIE DATASETS

CONCLUSION The moral of this story is simple. Quality social media data requires care, thought, and intelligence. Even when time pressures are not on our side and we need to complete our work quickly, quality results require significant efforts to be made for each and every dataset. The work may take several hours for a simple brand and several days of detailed, complicated work for a difficult brand but without care, the data can and will inadvertently offer incorrect results. You’ve seen the proof.

THE AUTHOR Annie Pettit is Chief Research Officer, Conversition Strategies, Canada.

LEVERAGING THE POWER OF SOCIAL MEDIA THE IMPACT OF SHARED NEWS CONTENT ON CONSUMERS Gregg Liebman • Carl Marci

INTRODUCTION The rise of social media over the last few years has been staggering. According to proprietary Nielsen data provided to CNN, Facebook alone created over 92 billion impressions in the third quarter of 2010. Forrester Research predicts it will be the “hottest” area of online advertising for several years to come. In addition, a recent survey of 104 national advertisers in 21 major industries resulted in 77% of marketers stating they will increase their spending on social media (figure 1). With this impressive growth and the number of large and diverse outlets ranging from social networks to blogs, it would seem that there is a plethora of readily available new advertising opportunities for marketers. However, a closer look at social media ad opportunities suggests that the ability to actually place engaging and effective advertising in what are often cluttered environments is limited, leaving marketers struggling to define their social media strategy.

FIGURE 1, SOCIAL MEDIA AD SPENDING IS SET TO EXPLODE Which of the following forms of non‐TV ads will you spend more on in 2010?

Social media 77%

Web advertising 73%

Search engine marketing 59%

Email marketing 46%

Marketing to mobile phones 46%

Event marketing 30%

Loyalty programs 27%

Word-of-mouth marketing 25%

Direct mail 20%

Point-of-sale marketing 20%

Traditional outdoor advertising 15%

Magazine advertising 13%

Radio advertising 11%

Newspaper advertising 7%

Other 4%

None of these 4%

Source: Forrester and the Association of National Advertisers (ANA); 104 national advertisers in 21 major industries

One strategy is to focus on social networks. These networks, which include Facebook, MySpace and Twitter, are websites that often get attention from traditional media sources and drive massive volumes of traffic. While social networks are a growing platform for advertising, with estimates of hitting close to $4 billion in 2012, up from $1.4 billion in 2009 (figure 2), there are questions about the effectiveness of ads placed in a social network environment (Beal, 2009).

As an example, compare two ad units running in Facebook and Twitter environments (figure 3). The ads are small and difficult to read as they are buried in a cluttered environment designed specifically so that audiences can move quickly between opportunities for social interaction. This raises an important question about whether social networks are destinations for content and information or a platform for social aggregation and the exchange of content and information from other sources. Is simply “placing” advertising in this environment the right approach or is there another more compelling model that taps into the inherent “social” component of these networks, including the sharing of content?

FIGURE 2, U.S. PAST AND PROJECTED SOCIAL NETWORK AD SPENDING GROWTH US Social Network Ad Spending, 2009‐2012 billions and % change

$3.93 (27.7%) $3.08 (55%)

$1.99 (39%) $1.43 (20.3%)

2009 2010 2011 2012 Note: Growth rate based on unrounded figures. Source: eMarketer, January 2011

FIGURE 3, EXAMPLES OF EXISTING ADVERTISING OPPORTUNITIES ON FACEBOOK AND TWITTER (RED BOXES)

One of the more frequently shared types of content online is news. A recent Pew Research Center study into news and social media suggests that online news has become a social currency -- a tangible product that people exchange with social value similar to offline goods and services. The results of the study show that 97% of consumers of online news also use social networks and 49% of social network users get their news online every day. The study concludes that the act of sharing news and conversation about news is an important and integral part of online exchanges in social media activity.

One of the most popular sources of news content shared online is CNN.com. With more fans than any other news provider and five times more followers on Twitter, CNN.com is the clear leader in shared news content (figure 4, source: Nielsen Buzzmetrics). In terms of shared news, it is interesting to note that the news site receives 14 million views per month of referral traffic from Facebook (second only to Google in terms of referral traffic to CNN.com). On a monthly basis there are nearly 850,000 comments posted by people about CNN news stories, often one or two word comments, but quite often paragraphs with insights, passions, fears, and feelings that people have about a story and they are clearly willing and interested in sharing to a broader audience.

FIGURE 4, CNN NEWS CONTENT SHARING ON FACEBOOK AND TWITTER COMPARED WITH OTHER MAJOR NEWS OUTLETS Buzz Buzz

CNN CNN

Fox News New York Times

Wall Street MSNBC News Journal

Yahoo News Fox News

New York Times MSNBC News

Wall Street Yahoo News Journal

CBS News Huffington Post

Huffington Post CBS News

0 10,000 20,000 30,000 0 200,000 400,000 600,000 800,000

Clearly, sharing news through social networks has become integrated, interactive and social. Importantly, from an advertising perspective, compared with Facebook and Twitter, CNN.com offers a broader palate for richer and more creative marketing communications (for an example, see figure 5). Despite the clear interest in sharing news and evidence that the palate for advertising was broader and more diverse, the question remained whether advertising related to shared or socially acquired news content had any unique impact on users. The goal of the present research study is to quantify the relative impact of socially acquired news and the impact of advertising associated with that shared news content.

FIGURE 5, EXAMPLE OF CNN.COM ADVERTISING PLATFORM FOR DROID

METHODOLOGY Neuromarketing is a rapidly growing sector of the research industry that applies principles of neuroscience and new technologies to market research practices in an effort to quantify and understand unconscious human responses. Biometrics, the application of biological sensors to understand brain activity, focuses on unconscious emotional responses that have a critical influence on human behavior -- an influence that has been elucidated with increasing regularity through academic books and papers from a wide variety of disciplines including psychology, economics, and philosophy. Methods

of research that can reliably measure unconscious emotional response have the ability to expand the sensitivity and accuracy of methods that once exclusively relied on self-report, which is more open to social and cognitive bias and more limited to conscious attributions of meaning.

CNN.com partnered with Innerscope Research, Inc., the only neuromarketing research firm to be accredited by a full ARF Research Review, and employed biometric methods for measuring unconscious emotional response to online news content and advertising that was socially acquired compared with the same news and advertising when it was not socially acquired. Biometric technology uses passive and non-invasive monitoring based on the physiological responses traveling from the brain’s emotional processing centers throughout the body via the autonomic nervous system (ANS). This allows media and marketing executives to learn and understand with incredible accuracy the unconscious emotional responses of their audience while that audience is actively consuming target media without interrupting the experience itself. Although people tend to find it easy to verbalize whether they like something or not (conscious attribution), they find it much more difficult to accurately explain “why” or “by how much” without relating the experience to other historical experiences or pre-existing biases, which vary tremendously from person to person. This process of “rationalizing” is a function of human evolution and neurobiology and it is an inherent limitation of traditional research methods.

Measuring unconscious emotional engagement allows researchers to eliminate the effect of cognitive biases from the research endeavor altogether. Measuring human biometric responses also allows for a common metric regardless of cultural background, language, religion, gender, or any other consciously imposed differentiation.

Emotional reactions are a complex combination of neurobiologically generated responses that are experienced in the ANS. These responses are measured because they are the unconscious building blocks of emotional states that guide attention, mark memory, and result in future behaviors. The clinically and academically validated biometric monitoring equipment wirelessly measures emotional responses by capturing channels of biometric data related to the ANS.

The research team for this study used unconscious biometric monitoring and eye tracking techniques to collect moment- to-moment data on each news story. Participants wore the Innerscope Biometric Monitoring System™ which is a chest- belt with advanced medical grade sensors that measure moment-to-moment changes in biologic indicators of emotion including skin conductance, heart rate fluctuations, breathing changes, and motion direction. These biometric signals are combined using a proprietary algorithm and are integrated with data obtained from eye tracking, which measures visual attention to the screen as well as visual attention to the brand and text. The final result is an index of unconscious visual attention and intensity of emotional response referred to as the level of biometric engagement.

STUDY DESIGN CNN.com and Innerscope set out to conduct an innovative study with the goal to understand, at a deeper level, the unconscious emotional responses that consumers have to socially acquired news and advertising in a social media environment.

Specific questions that were to be answered by this study were as follows:

 Question 1: Is content (news video, news articles and associated advertising) more emotionally engaging when it has been recommended by a friend and received via a social media website like Facebook than when that same content is randomly consumed outside of a social network experience?  Question 2: Is content (news video, news articles and associated advertising) more emotionally engaging when a person is in a state of recommending that content to friends?

The hypothesis was that people are highly influenced by their social relationships and will be influenced by experiencing news and advertising content that has been shared through a social network.

In the present study, 90 participants were monitored in three groups of 30 people each. Participants included friendship pairs of males and females, ages 25-54. They were regular users of CNN.com and Facebook. Social network theory claims that there are people who tend to gain a level of social status as people that other people want to listen to for advice or trends. Therefore, for the three groups, one friend was defined as an “influencer” using an academic model to define and screen for “influencers” (Patterson, Grenny, Maxfield, McMillan, and Switzler, 2008).

This innovative research design was utilized to replicate sending and receiving recommendations of CNN.com content via Facebook. Thus, the first group included the influencers who were in a “recommending” state of mind and chose which news items they would post recommendations for on Facebook. The second group included the recommendees who were open to and would receive the test content from their friends via Facebook. Finally, the third group was a control group,

matched for demographics they simply viewed the same test content in a neutral state of mind, in randomized order with no expectation to share and no indication of the source of the content. The study was conducted at the Innerscope Research Media Lab in Boston, Massachusetts in June, 2010.

In terms of the news content utilized in this study for the three groups, the target news items were chosen to be “evergreen” or timeless in nature, not overly dramatic, and light in both tone and seriousness. These items included four video and four static news items of various human interest stories include a sports highlight, a food/restaurant review, a movie review, a story on international refugees, and a book review.

The research team controlled for the news items seen by the recommendees in the second group because they could not control for which news items the influencers were going to recommend. Thus, the influencers chose the eight “evergreen” stories based on their own experience. All of the recommendees in the second group actually saw the same four news items being “recommended” by their friends in order to control for any differences in the quality or emotionality of the news content. This allowed the research team to control for the news content and focus on the power of social acquisition compared with non-social acquisition.

All participants wore the research agency’s biometric belt under their upper garments. The belt is applied around the chest easily in less than 10 minutes and requires minimal set-up time. All participants also experienced the content on a 19” high resolution color monitor with built in eye tracking (Tobii, USA). They all experienced a baseline media reel to establish response levels and timing for use in analyzing the target content. The influencers in the first group were given the eight CNN news items in a controlled online website that mirrored CNN.com. They were then instructed to post four pieces of content from the eight that they would recommend to their friend via Facebook.

Following the baseline experience, the recommendees in the second group were first shown the four news items that were chosen as the non-recommended content. Thus, the recommendees started out in a neutral state of mind for the first four pieces of content. Next, they were taken to their Facebook page where they saw that their friend in the other room had “recommended” four additional news stories. The four news items revealed to the recommendees were chosen by the research team to eliminate any influence of the stories themselves. There were two video and two static pieces of “recommended” content for participants to experience and the order was counterbalanced to control for order effects.

Finally, the control group experienced the baseline reel and then was exposed to all eight news items in a randomized order. The research team counterbalanced to control for order effects when relevant, but for the control group there was no Facebook interaction and no recommendations or “recommended state of mind.”

RESULTS Influencers, the participants in the first group who recommended the news items to their friends, were two times more engaged with the news items they chose to recommend compared with the control group that experienced the same news items in a neutral state of mind. Interestingly, an increased level of engagement was found with the advertising as well. The influencers in group one were 1.5 times more engaged with the advertising associated with the news items they recommended compared with the experience of the control group to the same advertising.

What about the recommendees in the second group? The effect was even larger for them. Recommendees, the participants who were introduced to half of the target news clips and associated advertising randomly and the other half via recommendations on their Facebook page while in a theoretically “receptive” state of mind, were 3.7x more engaged with the recommended news items than with the articles experienced randomly (figure 6). Interestingly, the effect was even larger for the advertising. The advertising associated with the target news items when recommended by a friend was five times more engaging than the advertising associated with the articles consumed without a prompt from a social network.

FIGURE 6, INCREASE LEVEL OF BIOMETRIC ENGAGEMENT WITH NEWS ITEMS (A) AND ADVERTISING (B) FROM FRIEND ON FACEBOOK A. B. Effect of Receiving Recommendation Effect of Receiving Recommendation Content Advertising Emotional Engagement Index Emotional Engagement Index

400 500 Recommended Content Reco mmend ed Ads (497) (370) 450 350

400 300 350 300 x 250 300 x 250 250 Ad Ad 300

200 250

200 150

Non‐Recommended 150 100 Content Non‐recommended Ads 100 (100) (100) 50 50

0 0

DISCUSSION CNN.com and Innerscope Research set out to understand the power of recommended news content and its impact, if any, on the associated advertising, in a socially sharing environment. The primary goal was to understand this dynamic from the perspective of the recommendee, the person who is on his or her Facebook page and receives an opportunity to experience a news item that is recommended by a friend. There are many more recommendees online than influencers so this was the primary objective of the study. CNN.com and Innerscope expected an impact on the level of engagement with this news content for the recommendee and found this to be the case. The magnitude of the effect was large and significant at 3.7 times the indexed level for the non-recommended content. While the fact that there was an effect was not surprising, the magnitude of the effect was to some degree. Even more surprising was the magnitude of the effect on the advertising. Participants experiencing advertising associated with the socially acquired news content were five times more engaged than they were with the advertising placed with the non-recommended articles.

This innovative study design also allowed for evaluation of the impact of news content and their associated advertising on the influencer -- those participants that were looking at content with a mindset and motivational state of sharing with a friend in their social network. Surprisingly, there was a two times increase in engagement with the news items and 1.5 times increase in engagement with the associated advertising that were recommended by the influencer compared with experiencing the same content without the power of any social expectation.

The magnitude of these effects is quite large. In the case of the recommendee experiencing the news and advertising associated with the target news items, the effect size would be the equivalent of going from the emotional intensity of a typical display ad online to an ad of significantly larger size or from a static ad that is “magically” transformed to a rich media ad by the power of social interaction. That is a large effect, especially considering that the CNN.com baseline for advertising is higher than a typical ad unit. These conclusions are informed by previous research that the research agency has conducted on CNN.com ad units as well as a baseline on rich media and video advertising units from other prior research.

One explanation of the magnitude of this effect is the power of social interaction on emotions. The results of the present study suggest that socially acquired news items offer a type of emotional transference or “halo” effect on the associated advertising that is experienced in a highly engaged state. Even in the case of video pre-roll advertising, the expectation of potentially passing along information via Facebook actually carried over to the advertising in the form of higher engagement for the influencers.

In summary, the results of this innovative study strongly suggest that when people are in an emotionally engaged state with one piece of content, supplemental content that is proximal or precedes it is gaining a benefit from the increase in emotional engagement. In this case, the proximal content that is benefitting is advertising. This emotional transference or “halo” phenomenon comes when the emotional response of one piece of emotionally engaging content is literally transferred over to another piece of adjacent content. Specifically, the present study indicates that the “halo” effect leads to more and higher levels of engagement with CNN.com news items and their associated advertising when the news item is recommended.

CNN.com has provided these insights to their ad sales units. They are in market selling both the “influencer” and the “recommendee” as valuable agents for the sharing of news as social currency via social networks. One of the great aspects of online advertising is the wealth of information that is available in order to understand users. While CNN.com has known for some time that many users take a story and share it with friends, make comments on it, or participate in a survey, the present study offers compelling evidence that these “social” users are also significantly more emotionally engaged with the news content they send and receive and are also more emotionally engaged with the advertising associated with it.

This research strongly suggests that the industry needs to redefine social media to include the content that is being shared in addition to the social networks themselves. If the social networks are the stations that aggregate and integrate individuals from a diversity of locations and backgrounds, then the locations of the shared content are examples of the destinations that are gaining increasing awareness and visibility as social networks grow. Individuals who come to CNN.com via Facebook, Twitter, LinkedIn, MySpace, and other social networks are experiencing CNN content through recommendations that have the additional emotional benefit transferred by the power of social interaction -- a power that works via unconscious emotional responses, directing attention and driving behaviors in increasingly complex ways.

REFERENCES Beal, A. (2009, April 24). Forrester predicts huge growth for social media marketing. Message posted to http://www.marketingpilgrim.com/2009/04/forrester-social-media-growth.html Pew Research Center (2010). Understanding the participatory news consumer: How internet and cell phone users have turned news into a social experience. Washington, DC: Purcell, K., Rainie, L., Mitchell, A., Rosenstiel, T., and Olmstead, K. Patterson, K., Grenny, J., Maxfield, D., McMillan, R., and Switzler, A. (2008). Influencer: The power to change everything. New York, NY: McGraw-Hill.

THE AUTHORS Gregg Liebman is Senior Vice President, CNN Ad Sales Research, CNN, United States.

Carl D. Marci is Co-Founder, CEO and Chief Science Officer, Innerscope Research, United States.

LESSONS FROM THE FRONT LINES HOW TO ENGAGE BRIC CONSUMERS IN MULTINATIONAL ONLINE COMMUNITIES Manila Austin • Katrina Lerman

INTRODUCTION The global marketplace is in flux, expanding rapidly to accommodate new markets and new consumers. As the use of the Internet and mobile devices explodes in developing nations, their citizens are drawn to the irresistible lure of the “global village” suddenly open in front of them. But as these consumers continue to grow their purchasing power and enter the digital world, they seek not only to connect with friends, family and like-minded individuals, but to gather information and inform their purchase decisions. Given that consumers with technology are likely also to be those with disposable income, it’s only natural that they would be interested in talking to brands online as well.

Our clients recognize the importance of understanding these markets, and are constantly pushing us to evolve, expand and perfect our international offering. At the 2009 ESOMAR Online conference, Communispace presented research examining emotional language in English-facilitated, multinational communities (Austin, 2009). The paper was well received, but it raised questions. Delegates asked us: How do you know you are hearing from the right people in different countries? What nuances are missed when you facilitate communities in English? How do you engage consumers in the rapidly developing countries of Brazil, Russia, India and China?

Two years ago we did not have a data-based response to these questions. This research is designed to provide some answers.

Currently, Communispace is facilitating over 175 online communities on behalf of more than 100 clients. Nearly one in eight of these are multinational, meaning they combine people from different nations but are facilitated in English (given its role as a “lingua franca,” particularly on the web). Anecdotally, we have noticed that some nationalities seem more difficult to engage, some language barriers harder to overcome (China, in particular, but also other developing countries); we have also observed that engagement varies from community to community and from client to client. We began to ask: How do community members from BRIC countries participate versus those from European or North American countries? What is driving these differences? And what can we do about it?

BRIC consumers and online participation in market research BRIC consumers are changing the digital landscape. Consumers in BRIC markets account for nearly one-third of the world’s online population and experiment with social media more broadly than any other consumers on the planet. They also represent the greatest potential for growth: in many cases, Internet penetration is still well under 50% (Piskorski and McCall, 2010). Market researchers who hope to use the Internet to attract, involve and learn from these consumers need to provide settings and formats that maximize engagement.

The ways BRIC consumers engage online are markedly different than those of people in more developed markets: for example, they experiment with social media (e.g., microblogging, social networking, media sharing) at much higher rates than consumers in North America, Europe and Australia. One explanation for this behavior is that they are joining a more diverse web, increasingly driven by mobile technologies, so these products and behaviors are native to them, rather than something to be adopted (Ibid).

Complicating matters for consumer insight professionals is the fact that budgets are not infinite, and conducting online research in English is often an economic necessity. Assumptions about language are particularly tricky when it comes to online communities, because – unlike panel surveys, which are one-way queries relatively straightforward to translate – communities are ongoing endeavors that depend on members talking to each other. For a true “community” to form (and not a series of sub-groups), all members, regardless of culture or nationality, must share a common language.

Given these constraints, recruiting members for this type of community presents a unique set of challenges. While BRIC and other developing countries may be more experimental and broad in their approach to online participation, the combination of low Internet penetration and English proficiency rates makes it challenging to find qualified members for multinational, English-facilitated online communities. Figure 1 shows an index composed of two data points by country: Internet penetration rates (Miniwatts Marketing Group, 2011) multiplied by scores on the English Proficiency Index (Education First, 2011)

FIGURE 1, THE INTERSECTION OF ENGLISH PROFICIENCY AND INTERNET PENETRATION BY COUNTRY

It is interesting to note that BRIC countries may present a challenge for researchers who seek to reach BRIC citizens online and engage them in English, but it is equally important to realize that these challenges are not necessarily true for other countries in Asia (such as Hong Kong and Singapore). And despite the commonly held notion that English proficiency is prevalent in India (due to its role in their education system) the results of Education First’s study indicate that India and China are nearly identical in terms of country-wide proficiency. Market researchers need to track these incidence rates when they are available, and anticipate the difficulty they will face in finding members from these countries.

We have certainly encountered these issues in our experience recruiting and facilitating multinational communities. To help ensure that clients find the most qualified possible members, we have found it necessary to conduct extensive screening for written language proficiency, including self-report questions and a hand-scored open-ended response. But language continues to be a challenge in the developing world, and we expect that it would contribute to differences in participation between countries, affecting both the quality and quantity of contributions.

These challenges are real, but they do not mean that engaging multinational community members—even in English—is impossible. Previous research suggests the opposite, in fact: Community members in global communities generally display robust participation, with cultural differences serving as a draw more than a barrier (Austin, Jennings, Lerman and Schlack, 2009). Additionally, because online communities are much more than research panels or brand fan sites (Jennings, 2011), they employ a range of activity types to elicit insight for the client sponsors and to entertain and engage members. We have observed members of our global communities engaging in the full range of activities available (e.g., discussions, surveys, product reviews, etc.) (Austin, Jennings, Lerman, and Schlack, 2008), and multinational community members have been shown to start their own activities at a higher rate than members of U.S. domestic communities (Ibid, and see also Lerman, 2008).

So BRIC countries are an important market, and although their consumers’ online behavior is complex it is necessary and possible to find, invite, and involve them in online communities. Market researchers who seek to do so, however, must provide settings that resonate with these web-savvy citizens. The methods and practices that have worked in western countries may not be the most appropriate to generalize across the globe. We need to re-examine our approach to engagement, and this work begins by focusing on the fundamentals of market research online community (MROC) theory and design.

MROC theory and design Because private, online communities provide longitudinal, 24/7 access to a small, dedicated group of consumers, they represent a very different model from traditional survey or focus group research. Indeed, they represent a unique endeavor specifically designed to help clients meet specific business objectives, whether tactical and feedback-driven or more exploratory and emotional. But the member experience is just as critical, for the value of the research is predicated on members returning to the community on an ongoing basis. Past research has shown that, beyond monetary incentives,

successful communities help meet various social, personal and professional needs (Schlack, Jennings, and Austin, 2007) of their members; for example, the need to feel heard or display expertise. Bringing these two forces – members’ social needs and company-sponsored formats for interacting online – together is the foundation for engagement and shared value. (See figure 2.)

FIGURE 2, THE SHARED VALUE PROPOSITION

Creating a culture and format that will deeply engage members requires the seamless integration of consumers’ social motives for participating and a company’s business reasons for inviting them in the first place. Every aspect of our consultancy, from recruitment to learning agenda to activity design to reporting, is centered on maximizing this shared value proposition. Depending on community composition and client objectives, this proposition can vary greatly, but it must be understood by all three parties (client, consumer and consultant). Our experience has shown that clients who derive the most value from their communities have a clear idea of why members are there and how best to take advantage of this ongoing, two-way dialogue.

There are two sources of information consumer insight professionals have available to help them monitor and maximize this shared value proposition. One is the extent to which we can observe and measure participation, which allows us to track engagement based on key behaviors. The other is the consulting relationship that exists between community managers and clients; it is those people who are in online communities every day, hosting activities and interacting with members and who are strategically working with clients who can provide us with ongoing information as to how well a community is accomplishing its purpose.

Objectives of the current study The objective of this research was to extract best practices to help online market researchers improve their ability to recruit and engage consumers in developing countries, with a focus on BRIC. In particular, we explored the following questions:

1. How does engagement vary by country? Based on our experience, we would expect participation to differ depending on country, with members from BRIC countries participating at lower rates than those from more developed or westernized countries. 2. How does participating in English (when it may be a second – or third — language) affect engagement? Language barriers could explain any country-by-country differences in participation; we would expect countries with lower English proficiency rates to write shorter posts and participate in activities that place less of a demand on writing in English. 3. How might specific conversation topics affect engagement by country? Previous research suggests that community commonalities can override cultural differences. The current research sought to explore this possibility with a large sample that included members from BRIC and other developing countries. 4. How do the unique characteristics of specific communities, themselves, affect engagement? Because the online communities in this study were custom designed to provide unique value to members and clients, we would expect the “engagement proposition” to vary from community to community; but to what degree and direction? This research explored how community properties and country intersect.

This paper includes the results and discussion of our quantitative examination of the above questions, and also the thematic results of interviews with facilitators of global communities; it concludes by providing lessons learned and implications for practitioners and future research.

METHODOLOGY Having established that purchase decisions are emotionally driven and that much of our shopping is habitual, The Marketing Store believes that there is a great opportunity for retailers and brands to develop stores and promotional communications that are better able to connect emotionally with their customers. They therefore approached BrainJuicer to run a pilot study to understand the emotional context of the shopper as they went about their shop. Specifically, through this piece of work, they set out to establish:

Sample Quantitative data for this study were largely archival, gathered from a selection of Communispace’s global online communities and representing six months (1 July – 31 December, 2010) of on-going participation in a variety of market research activities. Because our goal was to understand engagement in English-facilitated, multinational online settings, we focused on those communities that had members from multiple countries, preferably BRIC (n = 6,161 community members across 16 communities, representing 84 countries).

As previously discussed, effective communities are designed, not to represent a general population, but to deliver against a shared value proposition for members and companies. So while this sample reflected an amalgam of our clients’ business objectives and was not truly random, gender and most age groups were well represented: 50.6% were female (n = 3,116) and 45.7% were male (n = 2,816); and 32.0% were 25 years or younger (n = 1,970), 32.8% were 26 – 35 years (n = 2,022), 21.3% were 36 – 45 years (n = 1,313), and 12.0% were 46 years or older(n = 742). The sample also included members from B2C and B2B communities (n = 5,191 and n = 970, respectively) and communities representing 10 major industries (see table 1).

TABLE 1, SAMPLE BREAKOUT BY INDUSTRY

Total Retail (B2B) (B2B) Technology Apparel & Footwear Banking & Financial Services Beauty (B2B) Consumer Electronics Consumer Packaged Goods Food and Beverage Hospitality Travel and Publishing

Number of 2 1 1 2 3 1 1 2 2 1 16 Communities Number of 596 327 396 894 1183 424 322 454 1141 424 6161 Members From countries where 566 327 137 893 1092 374 260 318 1141 194 5,302 n ≥ 100 Australia - - - - 214 - 14 14 - - 242 Brazil 17 - - 100 25 - 3 10 - - 155 Canada - - - - 105 - 18 23 - - 146 China 31 - 3 100 15 52 33 10 365 - 609 France 58 - - 94 5 - 5 8 - 18 188 Germany 58 - - 99 11 52 11 7 - 37 275 Hong Kong - 158 ------158 India 73 - 30 96 57 - 5 43 322 - 626 Italy 22 - - 71 10 51 8 6 - 59 227 Mexico 5 - 25 - 29 52 4 2 172 - 289 Russia 72 - 34 97 7 - - 8 - - 218 Singapore - 169 - - 80 - 13 3 - - 265 South Africa - - - - 143 50 1 7 - - 201 Spain - - 13 78 3 - 7 13 - 21 135 Turkey - - 32 - 8 52 1 7 - 28 128 United Kingdom 91 - - 93 89 - 19 13 103 31 439 United States 139 - - 65 291 65 118 144 179 - 1,001

Measures and data analysis

Participation. For the six month period being examined, multi-country engagement was quantified according to three participation metrics:

1. Percent Weeks Active. The percentage of weeks a member was actively engaging (i.e., logging in and contributing) over the course of their membership; so how consistently they were they engaged. 2. Contributions per Active Week. The average number of contributions (e.g., responding to a survey, posting to a discussion board, commenting on an image gallery) made per member per active week; so how much did members do when they were engaged. 3. Likelihood of Lurking. The percentage of weeks a member was logged in but not contributing; so how often a member was passively reading as opposed to actively posting content, given that they had logged in.

We took two additional steps to clarify and improve our ability to interpret results. For the majority of participation analyses, we chose to exclude all countries that had fewer than 100 members, resulting in a working sample of 5302 people spanning 17 countries (see table 1). Using all three metrics to describe engagement for 16 communities and 17 countries also proved unwieldy. Thus we created an index to simplify our broad findings, and used the individual metrics to shed additional light on community dynamics on a case-by-case basis (see figure 3 for formulas for participation metrics and index).

FIGURE 3, PARTICIPATION METRICS AND INDEX

Measure Formula for Calculating

Number of Weeks a Member Logged In and Contributed Percent Weeks Active Number of Weeks of Community Membership

Number of Contributions Made During Membership Contributions per Active Week Weeks a Member Logged In and Contributed

Number of Weeks a Member Logged In and DID NOT Contribute Likelihood of Lurking Number of Weeks a Member Logged In

% Weeks Active * Contributions per Active Week Participation Index % Weeks Lurking

Effects of language. Because we screen community members for English proficiency on a pass/fail basis, it was not possible to examine language effects directly at the individual level. We were, however, able to indirectly assess the effects of language on engagement by observing how comfortable—or not—members were participating in English. We imported 295,821 open-ended contributions from the same six-month time period into an automated content coding and text analytics tool. Average word count was calculated, the assumption being that longer posts indicated greater comfort with English. In order to narrow our sample to a more manageable size, we included only those communities that had at least 50 members from BRIC countries combined; this resulted in a sample of 231,979 posts from 15 communities and 12 countries.

We also examined the extent to which countries were proportionately represented in four different activity types: Discussions (arguably the most demanding from a language standpoint), brainstorms (where members are asked to anonymously post just a few words or one sentence), image galleries (posts of images and comments), and surveys (the least text-dependent of all, and also anonymous). For this measure, posts by country were calculated as a percentage to represent each country’s observed contribution to the overall data set (see Austin, 2009, for more detail on this procedure). This percentage was compared to the actual distribution of members per country (representing the expected proportion of contributions), resulting in difference scores that allowed us to compare countries to see which were over- or under-represented.

Effects of conversation topic. Using the same sample of posts imported into our text analytics software, we applied a proprietary dictionary to all contributions from member-initiated discussions (n = 92,797 posts) to ascertain and quantify which conversation topics were most engaging for members; “top topics” were rank-ordered based on number of posts for country and community.

In-depth interviews with multinational community facilitators In addition to data from community members, we interviewed facilitators of multinational communities to extract “lessons from the front lines” for researchers and marketers alike. A total of 11 facilitators from eight global communities participated in video-recorded sessions. Facilitators were asked to reflect on their experience engaging members from different countries. In particular, we developed an interview protocol to explore challenges they have faced and to elicit specific examples of how they have maximized member engagement and ROI for sponsoring clients.

HOW ENGAGEMENT DIFFERS DEPENDING ON COUNTRY, LANGUAGE AND CONVERSATION TOPIC

Working across national borders: Differences by country Our first analysis was to test the fundamental assumption that engagement would vary by country. Online communities, however, violate many of the assumptions on which parametric statistics are based — cases are not randomly assigned and observations are not truly independent (as community members, by design, interact with one another over time). Additionally, preliminary analyses revealed that our sample was not normally distributed and that the variability for each country group was dissimilar. For these reasons we employed non-parametric statistical tests, but used them sparingly, and relied heavily on descriptive and qualitative analysis to help interpret our findings and draw conclusions.

We performed a series of Kruskal-Wallis Tests to explore engagement differences across countries. As expected, scores on our broadest measure (the participation index) varied significantly depending on what country members were from (Mdn = 3.0, X2= 199.73, df = 16, n = 3282, p = .000). This over-arching finding suggests that country does play a role in driving engagement, and an examination of the group medians and means illustrates the magnitude of this possible effect (see figure 4).

FIGURE 4, MEDIAN AND MEAN SCORES ON THE PARTICIPATION INDEX BY COUNTRY

As we might expect based on incidence of English proficiency and national Internet penetration rates, western countries such as the United States, the United Kingdom, Italy, Spain, and Germany, as well as “westernized” Asian countries like Singapore, show the highest overall engagement. The BRIC countries, as well as South Africa and Turkey, scored relatively low on this composite measure.

This high-level analysis revealed two interesting patterns. The first was the perhaps-surprising finding that Chinese members were relatively engaged for a developing country (scoring more similarly to French members than to Russians or South Africans, for example). It would be a mistake, then, to assume that all “developing” countries engage online in similar ways. It would also appear that a country’s economic development does not ensure robust engagement unilaterally. The second pattern revealed can be seen in the elevated mean scores, relative to the median, for every country. This suggests that there are varying participation patterns within countries, and that for those where the spread is especially dramatic (like Russia, Hong Kong and India), core groups of highly engaged members are responsible for the observed mean scores.

But how, precisely, can we characterize the behavior of engaged community members in different countries?

We analyzed scores on the three participation metrics, and Kruskal-Wallis Tests revealed significant differences by country for every one (see table 2); this suggests that what country community members are from played a role in how often they

were engaged throughout their membership, how much they did when they were engaged, and how likely they were to passively read versus post. TABLE 2, SUMMARY OF KRUSKAL-WALLIS RESULTS FOR THREE PARTICIPATION METRICS Contributions per Active Average % Weeks Active Average % Weeks Lurking Week Mdn =26.0% Mdn =11.5% Mdn =2.0 Country X2= 209.38 (16, n = 5302) *** X2= 56.17 (16, n = 4372) *** X2= 246.77 (16, n = 5083) ***

Community X2= 347.58 (15, n = 6161) *** X2= 157.03 (15, n = 5094) *** X2= 261.22 (15, n = 5919) ***

Age X2= 55.13 (3, n = 6047) *** X2= 14.32 (3, n = 4992) ** X2= 91.11 (3, n = 5805) ***

Gender NOT SIGNIFICANT X2= 4.64 (1, n = 4992) * X2= 4.50 (1, n = 5805) * *** p < .000; ** p < .01; * p < .05

In order to understand what behaviors were most descriptive of engagement in various countries, we created a box plot for each metric (see figure 5). Box plots allow us to observe a number of characteristics about how data are distributed. In this case, not only did the medians differ on each metric country-by-country (dark blue line on the charts), but there was variability within countries, as well (the light blue “boxes” represent the inter-quartile range—the middle 50% of the data points for each country).

FIGURE 5, BOX PLOTS OF COUNTRY SCORES FOR THREE PARTICIPATION METRICS Average % Weeks Active by Country

Average Number of Contributions per Active Week by Country

FIGURE 5, CONTINUED Average % Likelihood of Lurking by Country

Referring to appendix 1, Russian community members show the lowest median scores for percent weeks active and contributions per active week, but the plots also show that a solid fourth of all community members from Russia participate at higher levels. Looking at other BRIC countries, Brazil’s participation is the most similar to Russia’s, but Chinese and Indian members’ participation does not stand out as particularly low. In fact, members from China and India, on average, contribute more when they are engaged than do members from the United States or the United Kingdom. It is important to note that, even though significant country differences were found on the contribution per active week measure, the data points tended to cluster around the median for all countries (BRIC included). The range of average scores by country was quite narrow: a difference of only 0.80 contributions per active week (Turkish members contributed the most on average, at 2.85 contributions per week, and Russian members the lowest, at 2.05). This suggests that, once engaged, community members, regardless of country, generally contribute multiple times every week.

Perhaps more telling than countries’ ranking relative to one another on a given metric is how specific countries performed across all three metrics. We can note, for example, that a large proportion of BRIC community members spent their time passively reading versus actively contributing; high likelihood of lurking rates depresses scores on the participation index and can mask robust performance on the other two measures. In general, these data suggest that members from westernized countries (Singapore, United States, United Kingdom, Germany, Canada, etc.) are less likely to lurk than are members from less developed countries (Turkey, Russia, China, Hong Kong, etc.).

So while engagement varies between and within countries, it is possible to engage members in multinational communities in some form, and under some conditions. The question becomes, then, not if we can engage members across countries, but how.

Native tongue or lingua franca? The possible effects of language The high lurker rates observed for BRIC countries, where English proficiency is not wide-spread, suggested that language may be a barrier for engagement. If this were the case, it makes sense that members from countries with lower English proficiency rates would write relatively shorter posts. To explore this possibility, we conducted a Kruskal-Wallis Test to look for differences in word count by country. It was necessary, due to the large number of cases (over 200,000), to perform this analysis with a randomly selected subset of the total sample (25%); but even with the reduced number of cases results showed significant differences (Mdn = 10.0, X2= 1885.38, df = 11, n = 57,660, p = .000). This means that members tended to write longer or shorter posts depending on their nationality; and data were analyzed descriptively to understand which countries, specifically, were yielding lower counts (and were possibly more challenged by the demand of participating in English, see appendix 2).

Examination of the means and medians showed members from China, Germany, and Italy have the lowest average word count per post. This is misleading, however, as the type of activities in which members prefer to engage can affect how many words they are likely to use (e.g., discussions tend to require more words, but fewer words are needed to participate in galleries, brainstorms and surveys). We calculated average word count per activity by country to understand this effect (see figure 5).

FIGURE 5, AVERAGE WORD COUNT FOR FOUR ACTIVITY TYPES BY COUNTRY

As expected, we observed that the discussion format yielded the highest word counts and surveys the lowest. We also saw that community members contributed differently depending on the activity type. But can we attribute these differences to international English proficiency rates, alone? The secondary research we discussed earlier suggests not. Based on a combination of low Internet penetration rates and low English proficiency, we would have expected lower word counts for Russia than we observed. Similarly, we would have expected members from Italy, Germany and France to have higher word counts than members from Brazil. But this is not what the data showed.

Comparisons of post length by activity by country, however, did suggest that certain activity formats were more engaging for some members than others. As in our 2009 paper, we assessed how relatively under- or over-represented members from specific countries were across different activity types (see appendix 3); and as with our prior research, we found distinct country-by-country differences that were hard to interpret. When we considered the results of the activity type, word count and participation analyses together we could only conclude that engagement was most likely influenced, but not determined by nationality or language. For example, Russian members were under-represented in all four activity types and were the some of the least engaged according to our metrics, but we know from the word count analysis that the amount they wrote in discussions and brainstorms was on par with or more robust than people from many other countries. Word counts for Chinese members’ were some of the lowest observed for all activity types, but these members were better represented in discussions than Brazilians or Russians (or French or Germans), better represented in surveys than Mexicans, Singaporeans, Russians, and Indians, and scored in the “average” range on the participation metrics.

A few points of clarity did emerge, as contradictory as much of this seemed. First, the variability in behavior demonstrated that there was something that was engaging members from different countries in some ways, some of the time. Second, offering a variety of activity types seemed important and a relatively simple way to ensure that there would be something for everybody, regardless of country. Thirdly, and following from the first two points, there did not seem to be any one country that was definitively un-engage-able; every country we explored scored in the fair or moderate range on at least one measure. And fourth, despite intra-country participation variability, we could see that members from every country were participating in multiple activities during weeks they were actively engaged.

Engaging conversations: Analysis of member-generated content One factor that should encourage engagement and drive community vibrancy is the content of the conversations, themselves. We applied our proprietary text analytics dictionary to those open-ended posts that were generated as an output of member-initiated conversations — so those activities that fall outside the clients’ learning agenda and which members start and often facilitate themselves. Our goal was to understand if some topics would prove more engaging than others.

A frequency count of posts per topic allowed us to rank the top 20 conversation topics generated by community members, which are shown in table 3. When members in English-facilitated multinational communities talked among themselves, the topics they initiated most often focused on lifestyle: media and entertainment, family, socializing, travel, fashion, and the technology that supports them in these activities. A second theme in their conversation was that of trading advice — in general and specifically in regard to shopping, value and managing household expenses.

TABLE 3, TOP 20 TOPICS RAISED IN MEMBER-INITIATED CONVERSATION Rank Order Member-initiated Topics Number of Posts per Topic 1 Technology 9,768 2 Entertainment 9,505 3 Media 9,113 4 Shopping (in general) 9,022 5 Family (in general) 8,718 6 Value 8,295 7 Fashion 7,793 8 Household Finance 7,742 9 Family-Kids 6,343 10 Advice Giving and Getting 6,027 1 Technology 9,768 11 Food Preparation 6,005 12 Hospitality and Travel 5,391 13 Convenience 5,279 14 Behavior-Change 5,055 15 Socializing 4,978 16 Community 4,827 17 Researching 4,819 18 Chores 4,723 19 Emotion-Happiness 4,625 20 Fitness and Exercise 4,580

When we examined top topics by country (focusing on the top ten), we observed that these themes remained largely intact across various countries (see appendix 4). It may be that “lifestyle” subjects represent a fundamental value proposition for member engagement when combining many nationalities in one online setting.

We felt it would be a mistake, however, to assume that “universal” topics can provide sufficient value to continually engage members. As discussed, the success of communities depends upon the thoughtful development of a value proposition shared by both clients and their customers or consumers. The conversation within communities is often specific to that joint endeavor and theoretically should motivate members to participate and inspire them to initiate conversations that are truly their own, but also in line with the community’s purpose.

We explored this possibility by examining top topics by community, the results of which can be seen in appendix 4. While there was still some measure of overlap observed, individual communities, unlike countries, seemed to have a particular constellation of topics unique to them. We calculated the number of “top-ten topics” shared in common and found that, indeed, there was more variability across communities than there was across countries (see figure 7).

FIGURE 7, AVERAGE NUMBER OF TOP 10 TOPICS SHARED IN COMMON

As the tables in appendix 4 illustrate, the topics by community are less general and more specific to the particular properties of individual communities. For example, top topics for a B2B community of academic librarians included books, education and experience. These topics were unique to librarians and were not shared by beauty salespeople (facial cream and skin care were unique to them) or female runners (who initiated discussions to provide interpersonal support and share their aspirations).

Although it seems logical enough to assume that nationality, and even language, might be solely responsible for driving engagement in online settings, these analyses suggest otherwise. As researchers, we often must simplify our subject matter and, out of habit or necessity, reduce people to their demographic buckets because it makes it easier to study “them” if they are organized into tidy groups. While nationality and language constraints should not be ignored in research, it may be that, for online communities, it is the unique intersection of person and environment that can explain how and when people engage.

PERSON-IN-ENVIRONMENT: THE INTERSECTION OF COUNTRY AND COMMUNITY

What comes first, the country or the community? Trying to describe the dynamics of engagement in online communities is a bit like the “chicken or the egg” dilemma: it is hard to say which came first—the motivation to participate and connect, or the specific means and opportunity to do so. The particular intersection of members’ needs and corporate learning agendas is the foundation for community design. We wished to explore the extent to which engagement depended on the nature of communities themselves, with their unique constellation of members, goals and characteristics.

The results of Kruskal-Wallis Tests indicated that engagement did indeed vary across communities and for all three participation metrics (see table 2), which suggests that some multinational communities were more engaging than others, perhaps more successfully delivering against a unique shared value proposition. To better understand which members from which countries were participating (or not) in which communities, we calculated the average score on the three participation metrics for each country group within 16 communities. Results are arranged from highest to lowest participation by community across the column headings, and highest to lowest participation by country in the rows (see figures 8 and 9, and also appendix 5 for lurker rate); shaded boxes indicate above-average participation on a given metric.

FIGURE 8, INTERSECTION OF COUNTRY AND COMMUNITY FOR AVERAGE PERCENT WEEKS ACTIVE Average Percent Weeks Active

B2C B2C B2C B2C B2C Toys B2C Oral B2B B2C Travel B2C Food B2C B2C B2C B2C Banking B2B B2C Academic Sports Retail and Care Academic B2B Loyalty and Retail Electronics Motor Electronics and Women Apparel Total Country Publisher Apparel Youth Games Ear ly Publisher Technology Pr og ram Beverage Youth Men Oil Women Financial Beauty Me n Research Women (Agency) Mom Adopters Librarians Members Youth (Brand) Services Italy 45.2 52.1 43.7 46.3 38.8 29.6 44.5 33.0 41.7 United States 44.1 54.0 38.4 48.4 35.3 46.2 55.2 28.2 40.8 45.0 42.2 29.2 28.8 40.7 United Kingdom 48.4 40.3 41.4 36.8 46.2 17.3 24.3 47.6 41.5 35.4 17.9 40.1 Australia 48.9 41.2 48.2 30.4 50.8 34.8 39.4 Germany 32.3 38.4 29.2 50.6 47.9 53.6 38.8 24.5 38.5 39.4 Canada 34.1 39.6 35.0 33.6 54.1 19.8 39.3 Spain 37.4 65.3 24.0 32.0 33.0 25.6 39.6 39.6 37.6 China 47.3 56.3 47.6 48.9 70.0 42.9 39.1 6.7 34.8 28.4 36.1 Singapore 70.3 37.5 56.3 37.5 32.7 35.5 Mexico 26.0 32.0 47.3 38.1 35.3 20.2 23.2 33.2 India 52.4 39.8 45.8 49.0 24.6 37.0 21.6 28.1 29.8 22.4 18.0 32.9 France 23.3 47.5 44.6 34.4 42.5 43.0 19.1 22.4 32.5 Hong Kong 29.0 29.0 Brazil 34.5 42.7 32.2 5.5 18.1 4.7 30.3 27.1 South Africa 55.3 44.5 48.0 25.8 15.4 24.2 Turkey 67.0 43.4 30.2 56.0 12.4 18.0 23.8 Russia 67.3 29.0 27.1 10.3 19.6 31.4 4.0 20.8 Total 45.7 44.2 44.1 41.5 41.0 39.1 38.4 36.4 34.6 33.3 31.0 30.9 29.5 26.5 25.7 20.9 35.1

≥ Country Average ≥ 35.1 (sample average)

FIGURE 9, INTERSECTION OF COUNTRY AND COMMUNITY FOR AVERAGE NUMBER OF CONTRIBUTIONS PER WEEKS ACTIVE Average Number of Contributions per Weeks Active

B2C B2C B2C Oral B2C Food B2C Travel B2C B2C B2C Toys B2 C B2B B2 B Ba n kin g B2C B2 C B2C Retail Ca re B2B and B2C Loyalty Retail Sports and Academi c Academic Women and Electronics Electr onics Apparel Total Country Youth Ear ly Technology Beverage Motor Oil Program Youth Apparel Gam es Publisher Publisher Beauty Financial Women Men Men (Agency) Adopters Youth Members (Brand) Women Mom Research Libr aria ns Services Turkey 3.0 2.5 3.6 2.7 2.0 2.1 2. 8 Mexico 3.0 2.8 2.9 2.8 1.8 2.6 2.8 1.9 2. 8 Spain 3.8 2.7 2.5 2.5 2.6 2.0 1.7 2. 6 Australia 2.8 2.7 2.4 2.2 3.0 1.9 2. 5 Hong Kong 2.5 2. 5 Canada 2.8 2.5 2.5 1.8 1.7 2. 5 China 2.5 2.7 1.5 2.8 3.0 2.4 2.3 1.9 2.3 1.5 2.1 2. 4 India 2.8 2.6 2.7 1.9 2.4 2.2 2.1 2.5 2.3 1.7 1.8 2. 4 United Kingdom 2.7 2.8 2.8 3.1 2.4 2.2 2.0 2.1 2.1 2.3 1.5 2. 4 United States 2.6 3.0 2.5 2.6 2.3 2.3 2.4 2.9 2.4 2.4 2.0 2.1 1.8 2. 4 Italy 2.4 2.5 2.4 1.9 3.0 2.3 2.1 2.1 2. 4 Brazil 3.1 1.5 2.5 2.4 2.1 1.5 1.7 2. 3 Singapore 3.2 2.5 2.2 1.9 1.3 2. 3 South Africa 2.4 2.3 1.2 2.0 1.5 2. 3 Germany 2.4 2.0 2.2 2.3 2.3 2.2 2.1 2.0 2.1 2. 2 France 2.3 2.6 2.1 2.3 2.0 2.1 1.8 1.7 2. 1 Russia 1.5 2.5 2.2 2.2 1.4 2.3 1.7 2. 0 Total 2.7 2.7 2.6 2.5 2.5 2.5 2. 5 2.4 2.4 2.4 2.3 2.2 2.2 2.0 2.0 1.8 2. 4

≥ Country Average ≥ 2.39

Looking at the data as an intersection of country and community allows us to understand dynamics in a nuanced way. For example, we can clearly see that certain country groups within communities fare differently on different metrics. Chinese members in the Food and Beverage Youth community show some of the highest average weekly contribution per contributor rates, but are below average in regard to how often they are engaged throughout their membership. And the contradictory findings about Russian members begin to make sense: they perform below average in many communities, but are some of the top performers in the B2B community of women who sell beauty products.

Post hoc analyses: Age and gender These findings underscore how a compelling value proposition can overcome barriers to engagement. They also demonstrate larger participation trends based on individual-level variables: age and gender. For example, the two men’s communities in this study displayed below-average numbers of contributions per week, while those communities that had members engaged more consistently were often composed of older customers or consumers (e.g., the two academic communities of researchers and librarians). Members of youth communities tended to contribute more each week they were active, but participated less often.

Kruskal-Wallis Tests yielded significant results, indicating that engagement was dependent upon members’ age and gender (see table 2). Yet it was difficult to discern whether or not young people, for example, were more active contributors because of their age or because they were participating in a community where being with other young people was part of the unique value proposition. In other words, was it the person or the environment?

In order to determine the effects of the environment on the age-participation relationship, we compared scores on the participation metrics for members of Millennial communities with young members participating in mixed-age communities. The results of Kruskal-Wallis Tests revealed a significant difference for the average number of contributions per week (Mdn = 2.11, X2= 19.03, df = 1, n = 1570, p = .000), and an examination of the group means indicated that Millennial community members posted more content on a weekly basis than did young people in other community settings (see figure 10 for comparisons on all three metrics).

FIGURE 10, MILLENNIAL COMMUNITY MEMBERS PARTICIPATING IN MILLENNIAL VS. MIXED-AGE COMMUNITIES

Similar analyses were conducted to explore differences in participation by gender in mixed- versus same-sex communities. Kruskal-Wallis Tests showed that women tended to be active during a higher percentage of weeks when they were in a single-sex environment (Mdn = 26.0, X2= 5.57, df = 1, n = 3116, p = .02), but did not show significant differences in participation for the other two measures. Men, on the other hand, showed significant differences on all three measures: Percent Weeks Active (Mdn = 26.0, X2= 18.87, df = 1, n = 2816, p = .000); Contributions Per Active Week (Mdn = 2.0, X2= 12.58, df = 1, n = 2306, p = .000); and Likelihood of Lurking (Mdn = 12.5, X2= 4.02, df = 1, n = 2695, p = .05). An examination of group means showed that men were actually less engaged in single-sex environments—in all- male communities, men visited less often, they did less when they were participating and they were more likely to passively read (see figure 11).

FIGURE 11, MEN COMMUNITY MEMBERS PARTICIPATING IN SINGLE-SEX VS. MIXED-SEX COMMUNITIES

Results of quantitative analyses suggest that engagement in multinational online communities depends upon a combination of factors working together — mix of countries, comfort with writing in English, the community’s purpose and the demographic qualities of participating members. These factors are determined when the community is designed and, if properly implemented, serve to create that shared value proposition that is so compelling for members and clients alike. Then it is up to the community management team to ensure that the community does indeed deliver on both ends of this proposition, week in and week out.

COMMUNITY PROFILES Our quantitative analyses showed that members from developing countries were participating at low, but viable levels in English-language communities, but there seemed to be evidence that, in certain situations, their engagement was much stronger than the overall numbers would indicate. Due to the longitudinal, relationship-based nature of the research we conduct, numbers often do not tell the whole story, especially when it comes to participation. Community Managers play an integral role in shaping these communities, both as consultants to our clients (developing a research plan, writing and managing activities, distilling and reporting insights) and as facilitators, interacting with members every day.

We have had varying degrees of success running multinational, English-language private online communities, while facing the unique cultural, linguistic and logistical challenges these endeavors provide. The following three profiles will explore

the dynamics within certain types of communities, as well as provide specific examples of success engaging members from the BRIC countries. The groups we’ll focus on — Millennials, Early Adopters, and Professionals/B2B — represent consumers segments we have successfully recruited and engaged in dozens of domestic communities. Part of the adventure for us internally was to see how our experiences would compare as we sought to engage members in developing nations.

Millennials We currently manage several international communities made up entirely of young members, reflecting the broad segment “youth” or “millennial” and representing the ages 16 to 25 (although some members are slightly older). These communities represent the largest populations of members from the BRIC nations within our study, in part because young people in those countries are most likely to be online and in part because they are appealing demographics to our clients. Our multinational communities in this category also tend to comprise fewer countries (typically 5-10), with a focus on gaining a better understanding of a few key markets.

Our participation analyses show that young people are the least active age group overall – contributing less often and most likely to lurk. To hear our facilitators tell the story, though, these communities are incredibly exciting and vibrant places to be.

Our millennial communities often resemble a global meeting place, as young members thrive on the chance to interact with peers from around the world. While we might expect the value proposition for members to center on brand affinity and the “exclusivity” factor, we found that an equal, if not bigger, draw is the chance to learn about the lives of a diverse cohort. The shared experiences (e.g., school, dating, growing up) and interests (e.g., media, food, pop culture, fashion) of young adulthood provide ample fodder for discussion. Everything from school to dating to fashion to media to food to pop culture is on the table, and members are active in participating in – and even starting their own – discussions around these topics. Given members’ propensity to start their own activities and keep the conversation going, some facilitators describe these communities as essentially running themselves.

"One of our biggest assets is that we are a global community. It's just so surprising how much members like to learn about other cultures, and it's not top-of-the-mind things either, like, "Oh, you can drink beer at 16 in Germany?" We do a community-building series called "Ask a Member From ..." and we've done South Africa and Germany and Italy, and in the German discussion people were talking about the Berlin Wall and Adolf Hitler, and it was just, like, "Wow, these members really want to learn about other cultures." Community Manger, Youth Beverage Community

Facilitators have developed a broad range of strategies in order to leverage this natural social glue. One is to field activities that promote and celebrate the diversity found within the community. For example, the “Ask a Member From…” series, in which the members from one country are “deputized” as facilitators for the week and field questions from the community about life in their country, has been utilized successfully across many communities. Facilitators also incorporate culturally-specific holidays and events, through homepage announcements and light, community-building activities.

For non-native English speakers in general, but especially for younger members, it is important to keep activities short and simple, or risk losing their attention. Long-winded descriptions only cause confusion, while overlong surveys result in lower-quality responses. Videos, photo galleries, collages and other multimedia activities work well to overcome language barriers in all multinational communities, but especially for social media-savvy youth. Not only are they a fun, interactive change-of-pace for members, but they are full of the kind of nuance and emotion that can get lost without mastery of a language. Members love “showing off” their home country, be it footage of the World Cup from South Africa or a tour of the best food vendors at the local market.

"We've done a lot of video projects in the community and those actually have been very … we've gotten a lot more of the emotional component there. People on the video tend to be a little bit more fluent, but we have been able to get a lot of the rich emotional component from the Indian members and Brazilian members and some Chinese members. … A real strength of our approach is that we do do that deeper dive and more of the special projects, rather than just surveys and discussions, and you can get that from people who might have a little more challenge with the written word." Community Manger, Consumer Electronics Community

Although Millennial communities display lower participation rates overall, it is clear in talking with facilitators that the members who do participate do so at a high frequency and with enthusiasm. In particular, members from countries that commonly teach English in schools (India and Mexico were cited) are most active across activities, while those who are

less proficient prefer surveys but are less likely to participate in a free-form discussion where their writing (and any potential mistakes) can be seen by the entire community.

Participation spotlight: One community is unique in this study in that it is sponsored by an advertising agency rather than one company. So, while members do not have any particular shared brand affinity, they have the opportunity to talk to many global brands, from a variety of industries. Participation is excellent overall, but this community has been especially successful in engaging its young members from China. Both Chinese and Indian members participate at above-average levels in regards to all participation metrics and word count (see figure 12). Notably, this community’s overall lurker rate of 20% compares very favorably to the youth average of 33%, and even the cross-community average of 26%.

FIGURE 12, CHINESE AND INDIAN PARTICIPATION IN AGENCY COMMUNITY

Community Composition: China, India, Mexico, United States, United Kingdom Community Size: 500 members (100 from each country) Ages: 18 - 25

This significant reduction in the number of lurkers — a consistent problem with both young and BRIC members—may be a result of their twice-weekly fielding pattern, as well as a client that frequently needs to pose questions with very short turnaround times, resulting in activities fielded off-schedule, often with an incentive attached. It’s not unusual for a rush survey (or three!) to garner 200 responses in 48 hours, and the facilitators have learned that they can depend on members to come through under pressure. In describing the vibrancy of this community, one facilitator also noted the disproportionate amount of member-generated content – about 4:1 compared to client-driven activities. It would seem that the large number of research communities flowing through the community each week not only keeps members coming back, but inspires them to respond with ongoing conversation of their own.

Early adopters Another common type of community is one made up of people on the “leading edge” of a certain industry, as evidenced by their adoption of new products and trends, as well as passion for the category and influential behavior (e.g., blogging, user reviews). Generally, both the value proposition and social glue for members in these communities focus around the common interest in the product category and the ability to have a voice in shaping its future (although the membership may also represent a common demographic, e.g., moms, Boomers). These communities tend to be a bit broader in their geographic reach, as sponsoring clients are focusing on a more narrow set of consumers in each market.

Similar to the dynamics we saw in youth communities, members tended to engage around their cultural differences as much as the brand itself. Looking at a community of “leading edge” personal care consumers, which displayed some of the most consistently strong participation from BRIC members, we saw that the excitement over a global meeting place is not unique to Millennials. One facilitator described their member spotlight activities as a “tour around the world,” and noted about the diversity, “I think that that's the social glue ... the fact that these people are being exposed to members from all over the world in a way that … people only hope to travel that much.”

Though early adopter communities generally glean above-average levels of participation from members from developing nations, there is still evidence that not all members are truly English-proficient, in that they may read very well and be able to express basic ideas through the written word, but struggle to communicate emotion and nuance. The use of Google Translator to translate posts from members’ native language (which has been spotted across communities) confirms this trend and often results in confusing, though well-intentioned, answers. These less-proficient members often gravitate towards surveys, which feature closed-ended questions, and brainstorms, which are anonymous. If the community strays too far from what they are comfortable with, this may lead to decreased or discontinued participation.

"Initially [members from China] may be active when they are first recruited—the same thing with Latin America—whereas, for some reason, we do find that these members are the first ones to sort of “falls of the grid," while the European and North American and Australian members tend to stay in the community a bit longer. My guess is that that has to do something with the language barrier. So maybe they initially thought this was more of a community that was going to have surveys, whereas when there's a discussion or a special project and it requires a lot more language skill, that may be a reason that they shy away." Community Manger, Personal Care Community

Given that language remains a challenge, it’s not surprising that some of the most valuable insight for the sponsoring clients of early adopter communities has come out of our special projects—activities that take place predominantly outside of the community, such as keeping a diary, taking pictures or making a video. This kind of on-the-ground, in-the-moment research provides culturally-specific context, depth and access that can’t be gleaned from written answers alone. As category enthusiasts, members are more than happy to act as the eyes and ears of the company in their home country— sometimes even without being asked. In the case of our personal care client, they were actually able to acquire more accurate, current information about product selection and availability from the members than from their own vendors.

"One of the neat things that's come out of this community is actually awareness, for our client, that maybe they didn't know that a particular product was in a certain marketplace. The community members have talked about it and we've asked them to show examples and presented that information back to the client. So it's even been a source of, sort of, ‘Here's what's going on around the world that perhaps you didn't know about.’" Community Manger, Personal Care Community

Members will often engage in discussion about the latest products and brands in the category, asking each other what they’ve seen and heard locally. They also love to find out how their feedback is being used, so “closing the loop” with a follow-up from the client ensures that members feel heard and keep coming back.

Participation spotlight: One of our clients in the business of manufacturing consumer electronics has had success with two long-term multinational communities, one for men and one for women. In this case, we do see some differences in participation by gender, which have been observed in previous cross-community research (Schlack, 2011). Men tend to participate at lower levels than women do, particularly in terms of number of contributions and word count. In addition, women tend to display deeper member-to-member interaction, and discuss more personal topics, often related to motherhood and other uniquely “female” challenges.

FIGURE 13, WORD COUNT IN ALL-FEMALE COMMUNITY VS. OTHERS

Community Composition: Brazil, China, India, Russia, France, Germany, Italy, Spain, United Kingdom, United States Community Size: 400 members (even distribution among countries) Gender: Female

The comfort of a female-only environment may help explain why the BRIC members of this women’s community show higher-than-average participation, particularly when it comes to word count (see figure 13). Facilitators speculate that the community may “liberate” women from certain developing countries who would feel less comfortable discussing intimate needs, wants and aspiration in the presence of men. Just as naturally-forming communities of women have always been among of the most vibrant on the Internet, this may indeed prove to be a successful model for engaging female consumers from the developing world.

Professionals and B2B A third type of community engages members in their professional capacity — be it as a corporate decision-maker or as the end user of a product. Although technically considered B2C relationships, when consumers engage with a brand in their professional capacity (e.g., users of an academic database, business travellers), community dynamics — and, indeed, members’ motivations for participation — more closely resemble B2B than they do other consumer communities. Several of our clients have used multinational communities to engage corporate customers and professionals from dozens of countries, both developed and developing. Since the community’s purpose is usually to engage current customers rather than gauge a market at large, they tend to represent a small number of targeted individuals from a wide range of countries, rather than a large number of consumers from a few nations. In this study, 30 of the 32 nations with members in only one community were found in this group. As a result, professional communities were by far the most diverse geographically, representing 80 nations from all six continents.

In talking with facilitators, it is immediately clear that these communities represent a different model of engagement. Overall, their participation is on par with B2C communities, but the dynamics inside look very different. The social glue is based around their professional affiliation; multicultural diversity is still a draw, but when conversation centers on differences, they are related to the job rather than culture at large. In one community of scientific researchers, sponsored by a major publisher and information solutions provider, members balked at attempts to engage them in discussion around (what they perceived to be) “frivolous” personal topics. As highly-educated professionals in a global industry, they were accustomed to partnering with peers from around the world and couldn’t care less about archaic local holidays. But they relished the opportunity to discuss shared challenges in the research world and help shape products they use every day.

"We think that what unites members is that they're chemists, they're researchers, they're both going through the same process professionally to get this thing published— they just happen to be from different countries. The fact that there's a lot of diversity is a plus, because they do exchange and bounce ideas off one another. We have researchers from Iran, researchers in Israel in the same community, but I think that what really brings them together is that … they're all highly advanced in their fields and looking to get more out of this product." Community Manager, B2B Publishing Community

The value proposition for the individual member is not just about feeling heard; it’s about improving one’s job — and, by extension, one’s life — on a daily basis. These private communities offer a dual benefit: the opportunity to shape the direction of a product or service that you work with, as well as a secure space in which to network, give and get advice and share best practices with a worldwide network of peers.

This access is even more important to members in developing countries when it’s something that’s hard to come by in their profession. Within the researcher community, for example, members working in the developing world face a variety of political and economic barriers to funding, information access, cross-cultural collaboration and visibility within the research community at-large. The community allows them to discuss, and even begin to overcome, these barriers in a way that may not be possible in other venues. It fills a significant unmet need in their lives and provides value on many levels; unsurprisingly, they have been among the most active participants. From the client’s perspective, they know that some of the members in developing nations may not be able to afford their products yet, but they are interested in hearing their voice as “future customers.”

"They don't have research institutions as well-established as we do [in the U.S.], and so for them this is a great opportunity to network with like-minded professionals. … Scientists are under constant, constant pressure to receive funding and to publish, and to do that you need to collaborate ... and the funding bodies require multinational, cross-country collaboration … so it's very easy for Westerners to meet this criteria, and then you have members in developing countries (very much so from India) where they see this as an opportunity to collaborate, to get in there, to meet other people, and they take their membership to a new level." Community Manager, B2B Publishing Community

B2B and professional communities also provide a great reminder of the importance of a communicative and transparent client. Since members have such high involvement with the brand, offering to engage them in an advisory role can backfire if you do not back it up with action to let them know you are at least listening, if not acting.

Participation spotlight: One of our multinational B2B communities is comprised of professional women, generally under the age of 35, who sell beauty and personal care products. The members hail from a wide range of countries, are targeted to be savvy style leaders (“movers and shakers,” in the words of one facilitator), and tend to represent higher socioeconomic classes, even within developing nations. The community’s purpose was not only to support, track and inform these important customers, but to refine and roll-out a new platform targeted at young women.

Because these salespeople also represent end users, the client has engaged members in a diverse range of activities to try to uncover the aspirations and needs of this key demographic. One particularly successful activity grouped members into teams to write poetry inspired by the youth platform. Image galleries, scrapbooks and other non-textual means of expression were also utilized to explore notions of what beauty means and how the brand can help them achieve it. A client who’s “up for anything” gives facilitators the freedom to experiment with various forms of research to find the best fit for this unique group of women.

FIGURE 14, RUSSIAN PARTICIPATION IN B2B COMMUNITY

Community Composition: Chile, China, Colombia, Czech Republic, Greece. India, Indonesia, Mexico, Peru, Poland, Portugal, Russia, Spain, Sweden, Thailand, Turkey, Ukraine, Vietnam Community Size: 400 members Industry: Beauty/Sales

In a sense, this one community encompasses all of the demographics we have spotlighted: youth, product enthusiasts, women and businesspeople. However, we can’t claim some kind of additive effect; on the whole, this community displays average participation levels for B2B. But this proposition seems to hold unique appeal for members from Russia (consistently the least active country in the entire study), who considerably out-perform their average on all measure (see figure 14). By speaking to their professional ambition as well as their personal identity, the members have a clear purpose and they have responded with active participation.

LESSONS: HOW TO MAXIMIZE YOUR COMMUNITY ROI Throughout our conversations with facilitators, there were certain overarching themes and dynamics that seemed to span community type or composition. As more and more consumers in developing countries get online and grow their purchasing power, brands are increasingly eager to expand their research to include this evolving population. We’d like to share some of these more universal themes in the hope that they may provide guidance to others looking to engage consumers in similar forums.

Purpose, planning and recruitment  Know why you’re there—and they should, too. Time and time again, we’ve seen the importance of a client who has a focused idea of which consumers they want to talk to and what they hope to accomplish. Especially in the developing world, where recruitment costs are high, but the costs of not being there are higher, it is essential to target the right members for the right type of research. But the perfect member will do you no good if they are not engaged, which means providing a compelling environment and tangible rewards. Ensuring that the needs of both client and consumer are met is the job of an experienced community management team, but they cannot operate successfully without a clear mission.  Recognize the limitations of the format. If you are planning a multinational online community to be conducted in English, it is important to remember the vast inter-regional diversity found within large, developing nations such as India and China, and ensure that the sample size is sufficiently large to represent this heterogeneity. The fact remains that the BRIC countries still present a fairly limited population of fluent English writers; so if a truly representative, deep dive into culturally-specific behavior and attitude is the goal, this continues to be accomplished most effectively— at least for the time being—through in-language research.  Make sure to do thorough language screening. Currently, our language screening process involves self-report on proficiency and then one open-ended question, with each answer scored by a facilitator. Though this has worked well in general, there is clear evidence—especially with China—that we have members in our communities who are not truly proficient at writing in (or even reading or understanding) English. A more rigorous screening process could

help alleviate this problem, though it may not be cost-feasible in certain countries, especially when other criteria need to be met.  Find a partner on the ground. Especially if your organization does not have a strong presence in or experience with a certain country or region of the world, having a facilitator or consultant based in the members’ home country can help ease the transition. They can provide a resource to help with translation and misunderstandings, advise on cultural sensitivities and norms, and provide context for current and historical events. Depending on the nature of the partnership and the need, they could also help with sourcing members during recruitment.

Managing cultural differences and language barriers  Beware the American/Western lens. Ideally, the facilitation team would bring a diversity of backgrounds to the table, both to boost authenticity and to provide first-hand knowledge of cultural norms and experiences. Keep an eye out for references that betray a prejudice towards one part of the world (e.g., mentions of the seasons), and be aware of who is in your community and potential sensitivities (religion, politics, humor, etc.) — it is likely that many topics will remain “off limits.” It is up to facilitators to model culturally-sensitive behavior and ensure that all members follow suit.  Leave your assumptions at the door. There are so many popular misconceptions around the developing nations, it can be hard to distinguish fact from fiction. It is incorrect — and potentially costly — to assume that one market is very different from another, or even from your own. For example, it is widely assumed that the Indian population is much more fluent in English than those in China, due to its elevated status within the education system. The truth is that English proficiency is almost identical in the two countries and, given India’s very low rate of internet penetration, there is actually a greater chance of finding a “qualified” member in China than in India. If you don’t ensure that your research is free of these kinds of biases from the start, you risk leading your clients down a very slippery slope.  Keep it simple. With multinational communities, it is especially important to draft activities that contain clear, neutral language, free of idiom or slang. This helps to avoid misunderstandings and ensure that no one feels alienated by being a non-native English speaker. Creative activities that reference cultural touch points and have been very successful in domestic communities (e.g., a “Mad Libs” fill-in-the-blank discussion or a “Buy, Sell, Hold” activity based on the stock market) may result in more confusion than inspiration. Some communities have had success using images along with textual descriptions to get the point across and make absolutely sure everyone understands the activity. Never call out a member’s poor English and do not allow other members to do so; instead, gently probe for clarification (offline if necessary) until the confusion is resolved.  Utilize multimedia whenever possible. Depending on the capabilities of both your platform and your members, multimedia activities can provide insightful, non-written content in international communities. Because many members are not native English speakers, and may be less comfortable writing than with other forms of communication, we have had a lot of success with activities such as image galleries, videos, voicemails and special projects (e.g., collages, heat mapping, diaries). With added visual and aural cues, subtlety and nuance are communicated in a way that may get lost in writing, even when the member is not speaking perfect English. It provides a fun, interactive task for members and may help engage those who are less active in traditional activities.

Delivering on the value proposition  Find and cultivate social glue. A best practice for all of our communities, it is especially important for members from diverse cultural backgrounds to have some common ground to make the community a desirable and engaging place to visit. Examples include a common interest related to the sponsoring brand/industry (e.g., technology early adopters, runners), lifestage (e.g., moms, Millennials) or professional pursuit (e.g., researchers, librarians, beauty consultants). It may not always be apparent exactly how social glue will form, but it’s best to start with some type of commonality so that members have a reason to want to talk to each other, and not just the brand.  Deputize members to go outside of the community and report back on what they are seeing and hearing in their home market. One of the greatest assets of a multinational community is dedicated members who are willing to go “above and beyond.” With their help, you can stay ahead of local trends, evaluate the in-country competition, get unfiltered feedback on specific messaging and advertising, track product availability and selection.  Leverage the diversity. Members are far more intrigued than turned off by the mix of many countries in one community. They love the ability to hear the diverse perspectives and experiences of individuals with whom they would not otherwise interact. The “Ask a Member From…” series and other member spotlight activities have been hugely successful across communities. Topics that have come up range from the current (World Cup in South Africa) to the historical (Berlin Wall in Germany) to the brand-specific (What flavors do you have in China?) and beyond.  Plan to spend extra time facilitating this community. As we have seen, success engaging members varies greatly by community, and there is no underestimating the role facilitators play in creating a successful environment. Responding to member questions, reading carefully and probing for additional information or clarification, handling incentives and special projects, keeping an eye out for inappropriate or potentially offensive content, drafting clear and concise activities and member communication — these are all typical facilitation tasks, but require even more time and attention in an international community. Immerse yourself in the local culture as much as possible and keep up on

current events around the world. It will make you a better consultant to your client, ensure that you don’t appear out- of-touch to members — and you might just learn something!

CONCLUSION Previous research has shown that consumers in America are eager to connect with people online in safe, trusted spaces (Austin and Bennett, 2011). An examination of our multinational communities would imply that that same is true of consumers around the world. International, English-language communities work well for reaching a global marketplace and getting a basic understanding of consumers around the world and their varying perspectives, especially when there is a clear, mutual value proposition to benefit client and consumer alike.

Ultimately, we found that there are no “bad countries” or “bad members;” members from any developing nation, provided sufficient English proficiency, can be successfully engaged in a long-term, reciprocal relationship with a major brand through a private, online community. The single most important factor that emerged in predicting a member’s participation level was not their native tongue, country of origin, age, gender or any other individual characteristic; it was the specific community of which they were a part, which speaks to the unique environment and value that each one provides.

These communities don’t chug along from the start like we have come to expect with our domestic endeavors. There are slip-ups and starts; misunderstandings and breakthroughs. We are still finding our way, and we are learning. The value they have added to our clients’ business is a testament to both members’ and facilitators’ perseverance, enthusiasm and dedication.

REFERENCES Austin, M. S., and Bennett, G. I. (2011). Tech fast forward: Plug in to see the brighter side of life. Oglivy and Communispace whitepaper archived at www.communispacece.com/research. Austin, M. S. (2009). Cultural differences: Draw or barrier? Optimizing engagement in multinational online communities. Presented at ESOMAR Online Research Conference, Chicago, IL. Austin, M. S., Jennings, R. M., Lerman, K. J., and Schlack, J. W. (2009). Breakthroughs without borders: Generating cross- cultural insights in multinational communities. Communispace whitepaper archived at www.communispacece.com/research. Austin, M. S., Jennings, R. M., Lerman, K. J., and Schlack, J. W (2008). Generating customer Insights from the global village. Communispace whitepaper archived at www.communispacece.com/research. Education First (2011). EF EPI: Education First English Proficiency Index (www.ef.com/epi). Jennings, R. M. (2011). 'Like' me: The dynamics of public vs. private social media. Communispace whitepaper archived at www.communispacece.com/research. Lerman, K. J. (2008). Beyond the "Other" box: Giving Customers an independent voice in your community. Communispace whitepaper archived at www.communispacece.com/research. Miniwatts Marketing Group (2011). Internet world stats: Usage and population statistics (http://www.internetworldstats.com/top25.htm). Mikolaj Jan Piskorski and Tommy McCall (2010). Vision statement: Mapping the social internet. Harvard Business Review, July-August (http://hbr.org/2010/07). Schlack, J.W. (2011). The “64% Rule:” What Real Customer Engagement Looks Like. Communispace whitepaper archived at www.communispacece.com/research. Schlack, J. W., Jennings, R. M. and Austin, M. S. (2007). Meeting business needs by meeting social needs in small communities: Why size matters. Communispace whitepaper archived at www.communispacece.com/research.

THE AUTHORS Manila Austin is VP Research, Communispace, United States.

Katrina Lerman is Researcher, Communispace, United States.

APPENDIX 1: MEAN AND MEDIAN SCORES FOR PARTICIPATION METRICS AND INDEX (N = 5,302)

APPENDIX 2: WORD COUNT BY COUNTRY, AND BY SPECIFIC ACTIVITY TYPES

APPENDIX 3: PERCENT UNDER- AND OVER-REPRESENTATION BY COUNTRY IN FOUR ACTIVITY TYPES ACROSS ALL 16 COMMUNITIES

APPENDIX 4: TOP TOPICS SHARED IN COMMON BY COUNTRY AND COMMUNITY

APPENDIX 5: INTERSECTION OF COUNTRY AND COMMUNITY FOR PARTICIPATION METRICS

BREAKING THE MIRROR IDEAS FOR AN INSPIRING DIALOGUE Jon Rodriguez • Hoi Kee Wong

INTRODUCTION

Co-creation at Philips Design The use of co-creative research tools is a key principle within people research at Philips Design. From healthcare in hospital contexts (Parameswaran and Nino, 2008) to people’s living rooms, Philips Design puts people at the centre of the design process.

“Looking for personal meaning and experiences, Philips focuses on techniques that allow more openness and flexibility. People are perceived as participants rather than passive respondents.” (Rameckers and Un, 2005)

Use of online tools The use of online tools to further develop this approach offers attractive possibilities.

“Internet is no longer a platform for the presentation of content; it’s a global platform for collaboration” (Tapscott, 2006).

Online platforms increase the recruitment reach, enable granular participant recruitment, empower the participant to share their story at their own pace and reduce the cost of setting up global research. More importantly, these online platforms have the potential to increase the interaction and collaboration between organizations such as Philips and consumers. As Sense Worldwide puts it, ‘Co-creation is short for collaborative creation. It creates business value by employing the collective creativity, knowledge, experience, skills and enthusiasm of people from both inside and outside of a business’ (Sense Worldwide, 2009)

RESEARCH SET UP

Objectives Both case studies were executed within the context of the Philips Design Know How program. The objective of this program is to explore new design methodologies to improve Philips’ people focused design innovation process and to experiment with new tools and methods.

The specific objectives per project were:  Case study 1 Menu Menu: to engage with a community of participants in Spain and Germany in order to understand their current cooking rituals and behaviors;  Case study 2 Mirrortime: to engage with a community of young participants in Germany, United States, Brazil and China in order to understand their personal care rituals and identity across life stages.

Methodology  External agencies delegation  Online platforms and social media for data collection  Face to face collaborative sessions for data analysis, opportunity identification and synthesis  Multidisciplinary ideation workshops  Online feedback and co-creation for concepts

In collaboration with external agencies, we set up online platforms for each project. We used different design research techniques to entice the target groups to share their experiences via photo, video and/or text. The social media character of the platforms triggered people to share their opinions, their interests and lives in a way that they are familiar with. The key aspects to keep them engaged, regardless of the topic was fun, creativity and the challenge.

CASE STUDY 1: MENU MENU

Recruitment methods and platform set up Menu Menu Due to the general nature of the topic (food preparation), recruitment was done using an open website1) and harnessed by social media (Facebook and Twitter), recruiting around 15 participants in each country.

An ‘of the shelf’ platform (Ning) was tailored for the specific project needs. It was organized in 10 open challenges which were introduced by each community manager in each region daily. Participants had to answer the challenges by documenting their cooking behaviors in different manners. Participants could react to each-other’s entries, which further enriched the outcome. In order to further increase the sense of engagement internally in Philips with the community we opened the challenges to the Philips team as well increasing the overall sense of participation, community and transparency in the process. People in the organization could see consumer’s cooking behavior and consumers could see Philips employees’ behavior.

With the aim of attracting the right people to the exercise, we had to create the right environment for recruitment. We developed an open site outside the Philips corporate web environment to engage with people that were passionate about food and cooking but had little time to cook. The site featured a short video showing some examples of challenges participants would have to go through. Fun, surprise and challenge were the overarching themes of the page, which helped recruit an engaged group of participants both in Spain and Germany. (See figures 1 and 2.)

FIGURE 1, ONLINE RECRUITMENT PLATFORM

FIGURE 2, BOX TO KEEP PARTICIPANTS ENGAGED

We developed three platforms: one in Spanish for the Spanish community, one in German for the German community and a third one in English where Philips employees would take part in the same activities in English. All platforms were accessible for every participant in each community and this ensured cross fertilization and exchange of ideas and comments.

To keep the participants engaged, a box was sent to each participant by post at the start of the project. The box contained a set of envelopes with exercises, which were only allowed to be opened every day. A calendar magnet for the fridge was enclosed as well, where people had to note their eating behavior for the upcoming two weeks and it was also used as a positive reminder of the research project.

Research questions We spread the challenges to be carried out over a period of two weeks to ensure we filled all our knowledge gaps with each of the different challenges:  Day 1- Kitchen portrait: ‘Make pictures of your favorite tools in the kitchen and the forgotten tools’. With this exercise we wanted to understand the current use of kitchen tools and appliances in the different context. Why people like some tools compared to others and why some become the ‘forgotten tools’.  Day 2- Why do I eat what I eat? ‘Why do you cook what you cook and why do you eat what you eat?’ During this day we dig into the motivations of people in regards to eating. Participants share their views and the moderator triggers in depth in order to understand the motivations behind behaviors.  Day 4 – Pre Eating Post (PEP): ‘Share with us how you cook a meal from preparing the ingredients to serving on the table using only five tools’. We wanted to understand how people cook but also how people improvise when they lack tools.  Weekend - More or Less and Surprise guest: ‘Invite your friends for dinner and hand them an envelope, everybody will have a different request you will have to fulfill’. We prepared envelopes with different quotes like: ‘I’m vegetarian’ or ‘I am a diabetic’, forcing the participant to improvise different meals with different requirements for each guest.  Day 8 – Favorite food: ‘Invite your best friend for your favorite meal and he will ask you some questions’. We use the best friend as interviewer, preparing a simple interview guideline in the format of a paper tablecloth the friend can use to ask questions about the what and the how of the favorite dish.  Day 9 – 360: ‘SMS in the morning: make a 360 movie of your kitchen now and explain what you are doing’. We wanted to catch the participants by surprise in the morning and share with us their morning ritual. How is breakfast prepared on a Monday morning?  Day 10 – Calendar magnet: ‘Read your calendar of the last two weeks and share with us your thoughts’. Participants were asked to reflect with us what they ate during the last two weeks and what they liked and disliked most. This exercise also served us to reflect on the process and way of working with the participants.

Face to face collaborative sessions After two weeks of challenges it was clear that the community was looking forward to meeting each other. We responded to the request and organized a semi-formal lunch where we shared the main findings with them and had a good time together. Meeting the participants face to face increased the engagement of the team and helped participants share some very valuable granular insights for the project.

Rich insights After the online research, a multi-disciplinary design team collected the data from the platform in order to synthesize the key knowledge and develop inspiring tools for different opportunity and ideation workshops within Philips. We visualized a ‘cooking rituals experience flow’ which contained the key experiences of participants from gaining inspiration and shopping to cooking, serving and cleaning up. Using the experience flow, the team developed different people profiles according to different cooking motivations. In addition, we collected the most inspiring movies, pictures and quotes to be shared in the workshop. Different workshops took place in a professional kitchen environment, identifying insights, developing opportunities and generating new ideas. The team continued to share with the platform via video some snapshots of the workshops in order to keep the community alive and give some flavor of the ideas that were being generated. (See figures 3 and 4.)

FIGURE 4, EXPERIENCE FLOW OF COOKING RITUALS IN SPAIN AND GERMANY

FIGURE 5, OPPORTUNITY IDENTIFICATION WORKSHOP

Online feedback After the research phase, opportunity identification and ideation sessions, different concepts had been developed and articulated into one key image that would describe the key benefit of the idea. These concepts were shared with the community and different discussions were triggered in order to find out how the ideas resonated with the target group. In this way, we triggered the target group to co create and enrich the proposed concepts.

Due to the open character of the platform, participants built upon the concepts and on each other’s ideas. Having gone through the previous phase where we understood the cooking habits and motivations of the participants, it was very easy

to understand the reasoning behind some of the feedback coming from individuals. This made it much easier to understand and categorize the feedback in order to develop and enrich the concepts.

Conclusions This project has shown us how we can engage in very rich and inspiring conversations with lead users in specific topics that are of interest for them. The topic of food is inherently a high involvement and engagement topic which people are willing to talk and share about.

In addition this project tapped into fun and challenge as a driver for people to participate and increase their engagement and response. Proposing creative challenges brought much more creative responses. Putting more originality and effort in the way we asked questions paid back in the quality of responses received.

On the other hand opening the participation not only to the lead users outside the organization but including our own team in the research brought lead users and internal employees much closer, making the responses much more honest and candid and allowing us to obtain deeper and richer insights.

CASE STUDY 2: MIRRORTIME

Recruitment and platform set up In the Mirrortime project, recruitment of community managers and participants was done via the agency’s personal network and recruitment panel. A custom made platform by Sense Worldwide was used to create four communities on one platform (German, simplified Chinese, Portuguese, English). This enabled people to share their grooming experiences individually in a private task environment and to openly share their style and interests (style, girls, nightlife, YouTube movies, etc.) via a public blog in their mother tongue. The private tasks were equal for each community, but the public blog had its own organic growth and group dynamics. (See figure 5.)

FIGURE 5, ONGOING DIALOGUE BETWEEN PARTICIPANTS VIA PUBLIC BLOG

Each country was moderated by its own native community manager of a similar age and supervised by the agency and Philips Design. A group set up was chosen to encourage youth to talk freely in their own language and interest area and to stimulate them to share both their experiences and desires.

Research questions Over a period of 2.5 weeks, participants were asked to complete five private tasks and they were additionally challenged to keep the dialogue going on the blog by rewarding the ‘most expressive blogger’ with a flight ticket to the co-creation workshop in London at Sense World Wide.

Task 1: All about you – “Imagine a newspaper of 2015. If the main headline was about you, how do you think it would read?” In this task we want to find out who the participant is, what he looks like, what makes him proud and what are his key drivers in life.

Task 2: Your world – ‘Upload four photos of you right here right now and show us what you are wearing, where you are, who you’re with and what you are doing.’ In this task we want to grasp the world these youth live in, the people they hang out with and their normal look. Task 3: Looking good – ‘Back to the future, show us what you looked like two years ago and what would you want to change by creating a collage of your favourite style.’ In this task we want to understand what they find important in styling themselves and the look they want to achieve, what they like and dislike about their appearance by telling us what they wanted to change from their look two years ago, by creating a style collage and asking about their role models. Task 4: Getting the look - ‘Stop using your favourite personal care product for three days and tell us what impact it had on you’. In this task we want to discover their personal care routines and the products they need in order to achieve their style. Task 5: Mirrortime - “Show time: record a video of you shaving that you’d use to teach a younger brother or friend how to do it.” In the last task we want to know how their grooming experience has been so far and what is most important during this ritual. Face to face collaboration sessions Two weeks after the online research, one winner per country was selected to join the co creation workshop. The goal of the workshop was to synthesize raw online data with real people enrichment (four winners) into profiles, insights and experience frameworks.

Rich insights The profiles and experience framework from the co creation session were synthesized again by the research team in order to create visual workshop tools for internal opportunity and ideation workshops.

For each country, a life stage experience flow had been visualized to show the key milestones, identity, personal care transition and shaving experiences of teenagers, students and early career men. For each country, a multi-disciplinary design team created 12 personas (three per country) and was responsible for the personas throughout the whole project. In this way, cultural differences and similarities were maintained and challenged during the concept development. (See figure 6.)

Online feedback and co-creation To encourage the participants to talk freely about grooming, we selected 12 (three in each country) key participants from the previous online research to recruit six of their friends to join in the online co-creation session. In this way we recruited 12 girls and 50 boys to give feedback on proposed concepts. The mix of girls and boys resulted in interesting group dynamics and fun conversations. (See figure 7.)

FIGURE 6, SELECTION OF PERSONAS PER COUNTRY AND LIFESTAGE

FIGURE 6, CONTINUED, SELECTION OF PERSONAS PER COUNTRY AND LIFESTAGE

FIGURE 7, FINDING RESONANCE WITH USERS VIA ONLINE FEEDBACK DURING CONCEPT DEVELOPMENT

Conclusions By using online platforms, we have managed to research 60 youngsters around the world within their context and our given time constraints. By using an online platform, we narrowed the gap between this young target group and Philips, as we spoke a common language in a common space.

As grooming is a sensitive and unengaging topic, a fun and private approach suits well. Via the online platform, youngsters are stimulated to express their shaving rituals and desires, but also their uncertainties, doubts and dreams. In combination with a public discussion about style, we identified the main drivers and issues of their desires.

The online platform has given Philips the opportunity to speak and provoke young customers in a free, engaging and fun way and moreover has given the youth an opportunity to talk freely and honestly about the stuff they care about. This trust relationship and engagement has yielded rich, insightful and inspiring results for innovation within Philips Design and businesses.

FINAL CONCLUSIONS Based on these experiences, we have developed a set of principles that will help us articulate co-creation projects in the future. These principles will help us fully leverage the potential of this new paradigm, defining the way of working of a platform, our approach and tone of voice.

From anonymous respondents to community of participants In order for co-creation to be of value there should be open participation from the right stakeholders involved. If we truly want to create a sense of community, we have to break down the barrier between the corporate research body and participant to further the process of innovation. We have learned that the co-creation process is a self-enriching process, which is fueled by transparency and empathy. Therefore, not only research respondents should participate in the research challenges, where possible, we should also apply the research to ourselves. Sharing our own experiences with the external participants increases the sense of community between Philips and the outside world.

Transparency of the process creates trust between participants and increases engagement. Sharing the internal processes, design workshops and ideation sessions with the outside world makes the view of the corporate ivory tower much more approachable to participate and to engage with.

From interrogation to conversation Co-creation is much more than asking a question and getting an answer. The exercise of listening is as important as asking. Research plans need to be flexible and dynamic, adapting to the conversation and improvising plans on the go. It is not the same to talk to a busy mum about cooking as it is to talk to a teenage boy about his shaving experience, and you can’t use the same tone of voice for both of them.

From observation to provocation ‘Ask someone to cook a meal and you will see the obvious; ask someone to cook a meal with three ingredients and in five minutes and you will get inspiring results’. Or ‘Ask someone to show you how he shaves, will result in a chore; ask him to tell you how he would teach his younger brother to shave will reveal his underlying motives.’ In order for research to deliver inspiration it should stimulate inspiring situations. Propose challenging and extreme situations and you are more likely to tap into people’s curiosity and imagination. This approach requires taking the traditional way of defining research questions in a different manner. Rather than a rational exercise of question and answer, it requires imagination right from the beginning of the process. Balancing fun and challenge with the need to tap into the knowledge gaps will ensure richer and more honest responses.

ACKNOWLEDGEMENTS We would like to thank all the participants that made these projects a success because of the enthusiasm, honesty and energy they put into the process. Without their openness to share their experiences we would never have been able to carry out these projects.

We would also like to thank the Funky Projects team for their support in the Menu Menu project. Their creative and strategic thinking gave incredible energy to the project and team.

We also want to thank Sense Worldwide for their creativity in co-defining the research in the Mirrortime project and for their flexibility in developing their online platform according to our needs.

We would like to thank Juliana Vorstenbosch-Kelly for reviewing this paper.

FOOTNOTE 1. www.menumenu.org

REFERENCES Parameswaran L. and Nino, L. (2008). The healthcare journey: understanding clinical experiences from the patient and care giver perspectives. ESOMAR Conference on Global Healthcare, February 2008. Rameckers, L. and Un, S. (2006). People insights at the fuzzy front of innovation. Philips Design Tapscott, D. and Williams, Anthony D. (2006). Wikinomics. Sense Worldwide (2009). The Spirit of co-creation. THE AUTHORS Jon Rodriguez is Senior Research Consultant, Philips Design, Netherlands.

Hoi Kee Wong is Research Consultant, Philips Design, Netherlands.

THE EARTHQUAKE AND TSUNAMI IN JAPAN INVESTIGATE THE POSSIBILITIES AND LIMITS OF SOCIAL MEDIA RESEARCH FROM RESEARCHING HOW THE EARTHQUAKE CHANGED THE SENSE OF VALUE OF JAPANESE PEOPLE Koki Uchiyama • Michio Mutoh • Eriko Ota

INTRODUCTION On March 11, 2011 an earthquake struck off the coast of northern Japan, churning up a devastating tsunami that swept over cities and people. Recorded as 9.0 on the Richter scale, the earthquake was the most powerful ever to hit the country. The quake and tsunami resulted in more than 20,000 dead or missing individuals, and also triggered the worst nuclear emergency since Chernobyl; explosions occurred and radioactive substances were leaked into the air from the Fukushima Nuclear Power Station.

This disaster destroyed people’s daily lives that had until then been taken for granted. We think that it has changed consumers’ values and needs, which would lead to changes of purchase behavior as well as consumption behavior. It is important to capture such changes to make marketing strategies; however, it may be difficult to spot and analyze these new trends before the disaster took place by traditional polls or focus group interviews.

On the other hand, due to the spread of the social media, the new research method for collecting users’ comments and opinions from social media platforms and analyzing the market trend, the so-called “Social Media Research” has been introduced.

In Japan, the social media monitoring tool1) developed exclusively for marketing is prevailing, which allows not only to collect “data including preset key words posted in the social media” but also to gather social media data “exhaustively” on the web, including the past, and analyze target subjects retroactively.

Besides market research of goods and brands, reputation and brand strength research, and effective measurement research of campaigns and events, this tool has been applied in a variety of ways, from forecasting winners at a national election (80.2% prediction accuracy)2) to predicting the Nikkei Stock Average, with its advantage of maintaining mass data including the past extensively.

Social Media Research applied to assessing the reputation or brand strength of goods/brands and measurement of campaign/event effectiveness usually measures consumer’s “reactions.” However, users describe their lives in the raw in the social media, telling of daily occurrences and straightforward comments with background information, especially on blogs in Japan.

Therefore, Social Media Research can be a valuable method which enables us to extract not only the user’s’ “reactions” that come from a certain “stimulus” but also the “value” change which caused the “reaction”. However, we are not aware of a situation where the Social Media Research method attempted to catch the change in consumer’s “value”. And this is a very interesting area in which to explore the possibilities of Social Media Research. Thus, this paper presents the results of the survey we conveyed with Social Media Research to explore the signs of the Japanese people’s change of values triggered by the 3-11 earthquake and tsunami, and clarifies the possibilities and limits of Social Media Research evidenced from the survey.

This paper is organized as follows: first we explain the research data our survey targeted, the method of data extraction and processing, and the analysis method. Then, we describe our survey’s result and its analysis. Finally, we discuss the change in Japanese people’s values and clarify the possibilities and limits of Social Media Research.

SURVEY METHOD In our research, we decided to retrieve opinions and comments on blogs, not on other social media in Japan. In Japan, Twitter and Facebook is still new so that the users are considered as early adapters. On the other hand, blogs have already been used by a majority of people and there are more than 22.4 million blog users.3) In addition, the context of blogs is longer than that of other social media and the Japanese tend to write blogs as a diary. Therefore blogs in Japan tend to reflect consumers’ daily behavior and feelings in a straightforward fashion.

We carried out the survey as follows; 1) formulating hypotheses through teamwork; 2) extracting data from social media; 3) editing and processing data; and 4) analyzing results.

1. Formulating hypotheses First, through brainstorming we selected a set of key words which may reflect phenomena or values changed by the earthquake. We worked on this in groups composed of people from different backgrounds to exhaustively and creatively choose the words. Each member wrote down on sticky notes the items he/she thought represented changes before and after the quake, put them on the wall, categorized them, labeled the categories, and added the list of items by re-thinking from the labels. We repeated these steps.

2. Extracting data from social media We harvested data by using the Social Media Monitoring and Analysis Tool4) developed by Hottolink.5) This tool enables us to analyze posts on blogs, bulletin boards and Twitter. As of August 2011, 22.4 million bloggers were targeted and 1.88 billion blog posts collected during the period between November 2006 and August 2011. As this tool not only collects “data including preset key words posted in the social media”, harvests social media data “exhaustively” on the web (including the past), and allows us to analyze a target subject in real time, it is suitable to use it for our survey requiring data extraction from the past.

The period we set for data extraction was three years from August 1, 2008 to July 31, 2011. By comparing the volatility of pre- and post- periods of the quake with that of the same periods in the past two years, we examined whether the change in the blog posted numbers was caused by the quake or by a seasonal factor. . We found on the blogosphere a series of blogs exploited for affiliate spam. Spam blogs are created for promoting affiliate sites and their contents are auto-generated. These are noise for opinion retrieval. Therefore we detected and eliminated posts from spam blogs, utilizing the spam blog filtering technique installed in Hottolink’s Social Media Monitoring and Analysis Tool.

To search blog posts that express an opinion about the topics, we prepared a set of key words selected from the process of developing hypotheses. When setting the key words as the search condition, we took into consideration the language character in which specific words can be expressed in several ways with different usages and their combinations (hiragana, katakana, and Chinese characters). In Japanese there are several ways of writing one word.

Furthermore, in order to filter out unrelated phenomena, we also set a “not” criteria. For example, in Japanese “uchiage” means “drinking party” but it also means “skyrocket” for skyrocket fireworks, therefore, setting “fireworks” as a “not” keyword we were able to avoid extracting unrelated phenomena. In addition, to distinguish between wants and actual behaviors, we used key phrases such as “I want to XX,” “I want to buy XX,” “I want to go to XX” for the wants, and “I did XX”, “I bought XX”, “I went to XX” for the actual behavior.

3. Editing and processing data To improve the quality of the survey results, it is important to eliminate the elements contributing to the change of the blog post numbers, besides the influence of the quake, as much as possible. For example, when the number of blog posts about certain targeted phenomenon decreased after the quake, we can’t judge whether it means that a certain action was refrained from the quake or if it means that the total number of blog posts decreased. In addition, in the case in which the scope of data collection by Hottolink’s Social Media Monitoring and Analysis Tool was expanded and the blog post number relating to another targeted phenomenon increased, we cannot tell whether the phenomenon prevailed or the number of accessible bloggers increased.

Therefore, we normalized the daily blog post number containing the targeted phenomena divided by the entire blog numbers posted on that day in Japan.

Hereafter, “the number” in this study means the normalized blog posted number divided by the aggregated daily blog posted number and multiplied by 1,000, if not otherwise specified. This also applies to the graphics included in this paper.

In addition, to identify seasonal factors, we took the average number of blog posts in the period before the quake took place (from January 2 to March 4, hereafter referred to “the pre-quake period”), that in the period of that the quake took place (from March 5 to April 28, hereafter referred to “during the quake”) and the period after the quake took place (from May 1 to July 30, hereafter referred to “the post-quake period”) , and used it as supplemental data to identify seasonal factors.

4. Analyzing results We divided the surveyed phenomena into three subsets based on the charts of time-series variation and the change in numbers between the pre-quake and post-quake periods: the phenomena with increased numbers after the quake, the phenomena with decreased numbers after the quake, and the phenomena with numbers which remained the same after a temporary change.

SURVEY RESULTS AND ANALYSIS

Formulating hypotheses We listed 450 phenomena which we think reflect some changes since the March 2011 quake and tsunami, and divided these into 15 large categories and 54 intermediate categories. The results are summarized in table 1.

TABLE 1, CATEGORIES OF PHENOMENA LISTED AS REFLECTING THE IMPACT FROM THE QUAKE

2. Extracting data from social media We set the conditions to detect and extract the blog posts number for the 608 items identified in the process of generating hypotheses.

3. Processing data Elimination of the effect caused by the change of the monitoring machine’s data collecting scope Figure 1 shows that the number of collected blog posts rapidly increased by the end of March 2009. In figure 2, the 2009 line only skyrocketed at the end of March, 2009, in comparison with the other years, illustrating the same trend illustrated in figure 1. Figure 3 shows that the increase at the end of March 2009 was the same level with that of the other years.

FIGURE 1, TRANSITION IN THE NUMBER OF TOTAL BLOG ARTICLES COLLECTED

FIGURE 2, TRANSITION IN NUMBER OF BLOG POSTS CONTAINING THE TERMS “BEER” ALONG WITH “DRINK” OR “WANT” OR “BUY” /NON-NORMALIZED

FIGURE 3, TRANSITION IN NUMBER OF BLOG POSTS CONTAINING THE TERMS “BEER” ALONG WITH “DRINK” OR “WANT” OR “BUY”/ NORMALIZED

As seen above, we can eliminate the impact of the change of the monitoring machine’s data collecting scope by normalizing the extracted blog post number relating to the selected items by the aggregated daily blog posted number.

Elimination of the effect caused by a seasonal factor We would like to calculate the impact of a seasonal factor on the change of the blog post number by taking an example of “a desire for drinking beer”.

Now, let’s examine figure 3. It shows the number of the post-quake period increased from that of the pre-quake period. This trend line is, however, similar to that from August 1, 2008 to July 31, 2009, and that from August 1, 2009 to July 31, 2010, increasing in the post-quake period, in comparison with the pre-quake period.

This may be caused by a seasonal factor in which consumer demand for beer would increase as the summer approaches.

Looking at the change rate between the average number in the pre-quake period and the average number in the post- quake period, the change rate of 2009 is 1.435 and that of 2010 is 1.657, while that of 2011 is 1.515. This means that the variation of desire for beer in 2011 is within the normal range. In other words, the results of figure 3 and the comparison of the change rates lead to the conclusion that the desire for beer was not influenced by the quake.

As we attempted here, we can examine if the phenomena are influenced by the quake or not by comparing the change rate of the normalized blog posted number between the pre- quake and the post-quake periods and that of the same periods in previous years.

In our survey, for every phenomenon we identified as representing the change after the quake, we calculated its ratio obtained by dividing the 2011 change rate (A) by the 2009 change rate (B), and listed 40 keywords with largest change rate ratio as table 2, and the least 40 as figure 3. The ratio = (A)/(B) (A) The change rate of the average number between the pre-quake and the post-quake periods of 2011 (B) The change rate of the average number between the pre-quake and the post-quake periods of 2009

TABLE 2, LIST OF 40 KEYWORDS WITH LARGEST CHANGE RATE RATIO (THE 2011 CHANGE RATE / THE 2009 CHANGE RATE)

TABLE 3, LIST OF 40 KEYWORDS WITH LEAST CHANGE RATE RATIO (THE 2011 CHANGE RATE / THE 2009 CHANGE RATE)

Analyzing the results

The magnitude that the 3-11 disaster has had on the Japanese sentiment As the terms such as “quake”, “nuclear power station, nuclear power, and radioactivity” are closely related to the 3-11 disaster, it is reasonable that their number or change rates vary significantly.

In order to clarify how large the 3-11 quake exerted an influence on Japanese thinking, we made a comparison between the impact of the 3-11 quake and the Christchurch earthquake which took place on 22 February 2011, killing 181 people including 28 Japanese (hereafter, referred to as the Christchurch earthquake).

Figure 4 exhibits the distribution of the blog post number including “quake or aftershock”, setting the date of occurrence as 0. The maximum blog post number referring to the 3-11 quake was 45.1 times more than that referring to the Christchurch earthquake.

FIGURE 4, THE TIME-SERIES TRANSITION OF THE TERM “QUAKE” IN THE CHRISTCHURCH EARTHQUAKE AND THE 3-11 QUAKE

We anticipate that bloggers post their opinions and comments when they have strong sentiments to share with others. If it is true, the number posted on blogs may represent the major influence that the 3-11 quake had on Japanese thinking. The scale of an earthquake is measured by the Richter magnitude scale, indicating quantified physical forces. The blog posted number, however, can be a new indicator showing “the scale of impact in people’s minds”. We may conclude that the momentary impact of the 3-11 quake on Japanese was 40 times more than that of the Christchurch earthquake.

This indicator allows us to show the impact of various disasters - including an earthquake - on people’s mind as a common measurement.

Now, we would like to look at the transition of the number of the blog posts regarding “quake”, “nuclear power station” and “saving energy”, illustrated in figure 5 and figure 6.

The maximum blog posted number of these terms is as follows: “quake” 242 (on March 12), “nuclear power station” 50 (on March 15), and “saving energy” 141 (on March 14). On July 1, the date that limits on energy consumption came into effect, the number including “saving energy” is 2.6 times higher than that including “nuclear power station”. The accumulated blog-posted number of these terms from the day the quake occurred to July 31 is as follows: “quake” - 2,410; “nuclear power station” - 1,229; “saving energy” - 1,579.

FIGURE 5, TRANSITION IN NUMBER OF BLOG POSTS CONTAINING “QUAKE”, “NUCLEAR POWER STATION”, “SAVING ENERGY”

FIGURE 6, TRANSITION OF THE ACCUMULATED BLOG POST NUMBER REGARDING “QUAKE”, “NUCLEAR POWER STATION”, “SAVING ENERGY”

The results demonstrate that the power shortage triggered by the nuclear accident had a larger impact on Japanese people than the nuclear accident itself. Radioactive contamination has, indeed, generated fear throughout the nation, however, those under the direct and tangible danger at the level of having an effect on the human body are limited to the residents living within and around a 30-kilometer radius of the power station, while most have a concern about it but not as a serious imminent threat directly affecting our “ongoing” daily lives.

On the other hand, the power shortage has developed into a critical social issue for the whole nation, requiring us to change the way of both work styles (introduction of summer time, working shift change, cool biz allowing T-shirts and short pants) and lifestyles (setting air conditioners higher, reducing illumination, replacing power-saving home electronics). The power shortage is an “ongoing” problem having an enormous influence on our daily lives.

The signs of the value changes

Seeking healing away from fear The number of posts about “fear” on the blog sites has returned to normal from a rapid increase soon after the quake (figure 7). On the other hand, the words such as “hang in there (gambaro)” and “cheer up” have been increasing (figure 8).

FIGURE 7, TIME-SERIES TRANSITION FOR THE WORD “FEAR”

FIGURE 8, TIME-SERIES TRANSITION FOR THE WORDS “HANG IN THERE” AND “CHEER UP”

The decline in discussing the inward-looking lifestyle described as “goof off” and “withdrawal” may indicate that people have tried to take a more forward-looking perspective.

As can be observed from figure 9, the key word “healing” is also on an upward trend (figure 9). We assume that the people show a tendency to reduce their stress caused by fear and seek healing through some activities such as yoga, stretching and healing.

FIGURE 9, TIME-SERIES TRANSITION FOR THE WORD “HEALING”

Shift toward simple and modest consumption The key words such as “rich” and “make up” are on the decrease on the blog sites after the quake, while that of the word such as “refuse, disposal and separation (ideals of detachment from possessions)” are on the increase. The results may exhibit the people’s trend of a decreased materialistic orientation and growing awareness toward downsizing their lifestyle. Combined with the heightened consciousness of saving energy and ecology orientation due to the current power shortages, we see signs of an expanding tendency for “simple and modest consumption” among the Japanese after the quake.

Saving energy and reevaluation of traditional lifestyle Among the various measures for saving energy and insulating heat including “LED” and “green curtain (climbing plants over walls and windows)”, the number of the words “sudare“ (bamboo blind) and “suteteko“ (cotton knee-length pants) is on an increase, indicating that people are re-evaluating Japanese traditional housing styles and clothes and are adapting them for battle with the heat this summer, and respond to power shortfalls.

New perspective toward work style The increase of the words such as “summer time” and “cool biz” indicates that the response for power shortages have accelerated the tendency to seek efficiency and comfort on the job. The word “working from home” are on the rise (figure 10: Time-series transition for the word “working from home”), indicating a sign that people started to change their work style in the wake of the 3-11 disaster.

FIGURE 10, TIME-SERIES TRANSITION FOR THE WORD “WORKING FROM HOME”

New light on family bond The number of posts about “family” on the blog sites skyrocketed soon after the quake, but returned to the same level before the quake after a certain period (figure 11). We made a further attempt to observe the transition of the number of the posts containing the words we assumed showing relationship and community as much as possible, however we were not able to find data evidencing a new perspective toward the bond among family or people around.

FIGURE 11, TIME-SERIES TRANSITION FOR THE WORD “FAMILY”

A shift from growing consciousness toward social contribution to taking action The number of words such as “mutual cooperation” and “trust” has been on the increase (figure 12), showing the enhancement of solidarity/ community spirit.

The number of the words such as “volunteer” and “support” is on an upward trend, indicating the growing consciousness for social contribution.

FIGURE 12, TIME-SERIES TRANSITION FOR THE WORDS “TRUST” AND “MUTUAL COOPERATION”

FIGURE 13, TIME-SERIES TRANSITION FOR THE WORDS “CONTRIBUTION” AND “VOLUNTEER”

The number of the word “contribution” returned to the normal level with a temporary rise soon after the quake; however, the number of the word “volunteer” has maintained an upward trend (figure 13).

We assume that the desire to serve the public, triggered by the quake, was initially embodied as “social contribution” with the abstract idea that it was someone else’s problem; however, it has been gradually taken as “my” problem and common to be given shape as “volunteer”.

The people have shown more interest in social problems and we may see a change of their attitude toward taking more proactive stance.

CONCLUSION

A hint of the Japanese people’s value change The analysis of the opinions and comments posted on the blog sites demonstrates the following changes in thinking, behavior and values among Japanese consumers triggered by the 3-11 quake and tsunami.

 People initially tried to take a forward-looking perspective, getting out of an unstable situation soon after the quake. Going one step further, they have started to eliminate fear and have attempted to take actions for healing through some activities.

 The nuclear accident caused by the tsunami brought the possibility of power shortages over a widespread area centered in Tokyo. This has required the Japanese to change their lifestyle as well as work style. In terms of lifestyle, we see the sign of increased simple and modest consumption, reevaluating the traditional lifestyle.  In terms of work style, people started to demand more efficiency and we even see the sign of the change in their way of work such as working from home.

 The quake disaster has nurtured the awareness of solidarity/ community spirit and social contribution, which has driven people to take real actions shown as the words such as “helped” and “did voluntary work”, stepped up from the stage of desire/ awareness shown as the words such as “want to do some service” and “want to be a volunteer”. This altruistic behavior has become rooted among the people.

Lastly, we have grouped major changes before and after the quake in the tables below - table 4 for increased change and table 5 for decreased change.

TABLE 4, MAJOR INCREASE

TABLE 5, MAJOR DECREASE

The possibilities and limits of Social Media Research The possibilities of Social Media Research we identified through this survey are as follows:

 It enables us to extract the users’ change of awareness and behavior from the variation of and number of key words posted in the social media.

 By setting proper key words as a search condition, we can not only explore the change of users’ awareness and behavior, but capture the sign of their value change.

 It enables us to measure the magnitude of the impact that events including disasters have on the users’ mind by the variation and number of key words posted in the social media.

 It enables us to capture the change caused by a certain event by reviewing data before the event takes place. (Its ease in comparison with the past outperforms the traditional research method.)

 It enables us to examine massive hypotheses. It also allows additional review for hypotheses proposed in the middle of the survey without any trouble. (Its short research cycle for reviewing hypotheses outperforms the traditional research method.)

The limits of Social Media Research we found through this survey are as follows:  We cannot capture any change of awareness or behavior without setting proper key words under specific hypotheses.

 The extensive experience and knowledge would be essential to select appropriate key words. (In this research, we built a hypothesis that people would give more weight to “family” and “community”, but we were not able to set proper key words extracting such an orientation in their awareness or behavior.)

In terms of “family” and “community” orientation, the results of the traditional online questionnaire research are not consistent to show a certain direction of change.

Case 1: According to the results of the online research, “the change of the relationships in the wake of the earthquake disaster” (Dentsu Soken, 2011), to the question “do you have a relationship you want to cherish more than ever in the wake of the 3-11 quake?”, 74% replied “yes”.

Case 2: To the question “suppose you have another two hours on a weekday, who do you want to spend this with?” in the lifestyle survey carried out by Kirin Holdings, there was little difference between the replies of the December 2010 survey and those of the July 2011 survey and we were not able to capture a new perspective that people would hope to spend more time with family and friends after the quake (table 6).

TABLE 6, MAJOR DECREASE

The results of these surveys illustrate that when a questionnaire is prepared before an event takes place and it asks some change of awareness or behavior in the wake of the event, it can be a leading question, as in Case 1.

The future development of the Social Media Research method In this survey, linkage among the changed phenomena was provided manually in the process of our analysis. It is possible, however, to apply a certain hypothesis automatically for determining the linkage among the changed phenomena by performing a time-series correlation analysis on the number of the words relating to the changed phenomena. The correlation analysis, we suppose, could contribute to figuring out what change of “awareness” causes a certain change of “behavior”.

We only attempted a quantitative analysis in this survey, but by conducting a qualitative analysis on comments indicating behavioral change, it may be possible to capture users’ “awareness” and “value”, which trigger their behavior, in a more direct manner. These are the areas we expect for the further development.

Integration of traditional research method and Social Media Research Based on the features of Social Media Research this survey clarified, we would like to propose a research process combining traditional research method and Social Media Research.

The integrated method would work as follows: First, tracking the change after a certain event takes place in real time by using the method of Social Media Research we adopted for our survey, and formulating hypotheses, screening questionnaires and choosing proper terminology to carry

out traditional research, and subsequently, examining the selected hypotheses and questionnaires utilizing a traditional research method.

We think the new research process raised above would expand the possibilities of Social Media Research.

Another idea for the integrated process would be to explore the context which triggered some changes or their future impact by MROC (Market Research on Community), after building hypotheses and extracting agendas by Social Media Research described above.

FOOTNOTES 1. https://www.facebook.com/video/video.php?v=10150563971165454 https://www.facebook.com/video/video.php?v=10150563976940454 2. http://senkyo.kakaricho.jp/2009/country.html 3. Measured by Hottolink Inc. 4. https://www.facebook.com/video/video.php?v=10150563971165454 https://www.facebook.com/video/video.php?v=10150563976940454 5. https://www.facebook.com/hottolink.inc?sk=info

REFERENCES Dentsu Soken. (2011). “the change of the relationships in the wake of the earthquake disaster”: http://www.dentsu.co.jp/news/release/2011/pdf/2011083-0715.pdf

THE AUTHORS Koki Uchiyama is C.E.O, Hottolink Inc., Japan.

Michio Mutoh is Senior Planner, Business Development, Hottolink Inc., Japan.

Eriko Ota is Head, Kirin Institute of Food and Lifestyle, Kirin Holdings Company, Japan.

PART 4: MOBILE RESEARCH DIMENSION

DIGITAL MATRIMONY MARRIAGES THAT ARE TRANSFORMING THE FACE OF RESEARCH Kristin Luck

INTRODUCTION As a planet, we’ve become people on the move – with millions connecting on the fly through social media outlets and mobile devices ranging from smart phones to tablets. In fact, 2011 is widely being hailed as “The Year of the Tablet” and Forrester Research predicts that this year, tablet sales in the U.S. will double and that by 2015, nearly a quarter of all personal computing devices will be a tablet.

This rapidly changing section of the consumer market has impacted market researchers as mobile and social media research technologies and methods must be developed at nearly the same speed as new mobile devices. This presents several challenges to us as market researchers:

 How can we use technology to our advantage to connect, engage and deliver actionable results in the mobile universe?  How do we develop best practices for research design and implementation utilizing social media and mobile platforms?  What issues do we need to be aware of as we embrace these new research platforms and dive into technology development?

How to address these issues from a research perspective means we have to first understand how technology is changing the way marketers communicate with consumers.

WHY IT’S IMPORTANT … WE’RE ON THE MOVE! Mobile devices create an expectation for nomadic access to everything and anything we may want or need. The increased demand to be met on the move has changed the face of communication, and is swiftly changing the face of market research. Respondents expect to be met on their own turf - through their mobile phone or device.

With the advent of more robust mobile devices such as the iPad, which sold more than three million units in its first 80 days on market, reaching people wherever they are has become even more possible, and more important. iPad aside, the International Telecommunication Union recently reported there were five million mobile subscribers worldwide at the end of 2010. The ubiquity of mobile phones, their presence in the daily lives of users, advances in the wireless networks and improved handset technology, are all reasons that marketers are becoming hooked on the idea of mobile marketing. In fact, mobile marketing is one of the fastest growth areas in media adoption.

Mobile marketing and advertising will grow almost ten-fold over the next five years, from $1.5 billion to $11.5 billion according to eMarketer.

In addition, many of the top social networking sites have demonstrated consistent rapid growth in their global user bases. Facebook.com, the global leader among social networking sites, now has more than 600 million active users, and 30% of those users are accessing the site through their mobile devices. Those who access the site from a mobile device are twice as active as those who do not. This is a key indicator in the power of mobile in capturing an active audience.

THE INITIAL DRIVER … MOBILE MARKETING Mobile marketing refers to marketing campaigns delivered via text messaging (SMS), multimedia messaging (MMS) or wireless application protocol (WAP or what most of us think of as mobile web browsing). Mobile marketing can also be delivered via keywords or shortcodes (for example….text DEC [keyword] to 6622 [shortcode]…if you’ve ever been to a professional sports game you’ve likely seen an example of this on the JumboTron).

Even if you’re not receiving marketing communications on your cell phone (yet!), you can be assured that mobile marketing is here to stay. More than 40% of U.S. mobile subscribers say they have been exposed to one or more forms of advertising/marketing on their cell phone in the past 30 days. Not convinced? Ninety percent of Americans over the age of 13 own a mobile phone. There are 66 million SmartPhone users in the United States. Sixty percent of U.S. marketers are engaged in mobile marketing. Forty-five percent of major U.S. brands are experimenting with mobile campaigns and 40% of the 400 billion annual global text messages are commercial (Source:eMarketer).

WHAT IT MEANS FOR US AS RESEARCHERS As an industry we continue to struggle with decreases in respondent cooperation rates. Quality concerns continue to plague us and although sample suppliers have taken the brunt of the blame, there’s no question that survey design and length present major barriers to respondent retention and participation.

In addition, respondents are using Smartphones more frequently than ever to access online content. In a recent look at the tens of thousands of respondents that flow through our surveys at Decipher in a given month, over 20% of respondents were attempting to access through a mobile device. This provides a clear signal that providing long, involved surveys in a strictly online environment is quickly on its way to impacting respondent engagement as survey takers become more active on their mobile devices.

Research technologies must evolve. Online research needs a “spouse”. Our challenge is to not only keep respondents engaged but to connect with them using relevant communication modes.

THE SOCIAL MEDIA OPPORTUNITY The use of social networks has grown astronomically (nearly 25% worldwide, annually according to comScore data). Online social networking is no longer a “Gen-Y” phenomenon. In 2009 alone, 72% of Facebook users were over the age of 24 and every six months the average age of a Facebook user was increasing by one year. Since its inception in 2005, social networking has surpassed email as the primary mode of online communication.

So how does this impact market research? Panel response rates are declining. Email no longer works as a mode to recruit certain populations for research studies (Gen Y, ethnic minorities, etc.). Our pool of online respondents has moved on. How do we engage with them in their new natural environment?

In January 2009, Kim Dedeker (then at P&G) commented that we need to focus on “reconnecting with the consumer in their backyard”. The challenge becomes connecting with people in their digital backyard – and how best to reach them when you’re working through sites like Facebook, FourSquare, Pandora, and a myriad of other sites.

Peanut Labs became the first market research provider to really wade into the social media space and they did so by really learning to play in the social network monetization market. This market is fueled by direct marketing (“offers”) and advertising that allow social network users to earn currency by responding to commercial offers (sound familiar?). In essence, if research wants to access people in their new “backyard” we have to learn to play in the monetization market.

Social media provides us with the opportunity for the following:

 Accessing respondents in their new “natural backyard”  Achievement of higher response rates  Access to difficult to reach audiences  The ability to combine the principles of panel and river sample.  The creation of new data collection tools like widgets and “gamified” surveys that allow for a more enjoyable survey taking experience.

As an industry we have the opportunity to become a powerful source of monetization for social networks while improving on our existing recruitment and data collection methods.

Today, we’re using social media for both active and passive data collection. From sampling to data mining/text analytics to survey widgets the opportunity for harnessing social media as part of our research practice is limitless.

THE MOBILE RESEARCH OPPORTUNITY Mobile devices and social networking sites (society’s current word-of-mouth marketing) have the potential of being the ultimate relationship tool: these modes are personal, deliver a one-to-one communication with mass market efficiencies and are capable of distributing timely messages.

With the explosion of online and mobile media and entertainment options, static online surveys struggle to be effective at engaging respondents. New tools and data collection techniques are evolving to more effectively connect with respondents via interactive surveys, mobile surveys and social networking sites.

Mobile is without a doubt the fastest growing area of media adoption. Mobile offers us a direct connection (one to one!) with respondents and gives us exposure to the growing number of individuals who no longer have a landline. More than 90% of mobile devices are now mobile internet ready and 60% of Smartphone users access the web at least one time a day. Despite the perception that mobile phones can only be used for one or two question surveys, marketers today are conducting research via mobile devices for all types of research: event research, ethnography/diary studies, on-site customer satisfaction surveys, screenings and exits … the list is growing daily.

Despite the fact that mobile research is still in its infancy stage, surveys conducted via text messaging are already scarce with industry focus shifting to surveys conducted via WAP (mobile internet browser much like taking a survey online) or via a survey application designed for a specific site (think Facebook survey widgets) or phone operating system like an iPhone or Android device.

Both WAP (or web-based mobile surveys) and app-based surveys have their own benefits and challenges. WAP surveys allow for text and multimedia survey cross platform (meaning they’re compatible with mobile browsers on multiple operating systems). Device compatibility is 70% plus. The downside is that mobile browser speed can vary considerably based on the wireless connection. App-based surveys are device specific (meaning an iPhone app won’t work on an Android phone so multiple versions of the app are necessary to allow for cross platform research) but generally bring faster delivery and upload times. This may ultimately work to increase respondent satisfaction with the survey taking process. In addition, survey apps can be developed on and integrated into pre-existing apps which may present marketers with opportunities to add in survey functionality to apps that have served other functions to date.

Regardless of the TYPE of mobile survey, there’s been much industry concern about blending mobile and non-mobile respondents together in the same data set.

Recently Decipher conducted a comparative analysis between mobile and non-mobile survey takers. This data comes from two satisfaction trackers, which ran in field through July 2011 and neither of which were optimized or targeted for mobile users. Both studies had a similar construction:

 Roughly 10 questions  Mostly single select questions with a couple of grids  A few multi-select questions  One open-end question

In these particular surveys, under 10% of all click-throughs came from mobile users.

Overall we found the following:

 Mobile users take a little longer to get through the survey (no small surprise!)  Most mobile users are taking surveys via a Smartphone (namely iPhone or Android) or a tablet PC (such as an iPad).  Not surprisingly, non-mobile users give longer answers to open-ended questions. However, that difference was an increase of about one additional sentence. We found that some non-mobile users will give VERY long answers, while the most we saw from mobile users was two to three sentences.  Satisfaction ratings showed no clear differences between non-mobile and mobile users.

Mobile survey takers are clearly here to stay. Although just under 10% tried to access the surveys in our study we expect that to grow substantially and in a very short period of time.

THE IPAD OPPORTUNITY Although we’re lumping iPads in with mobile devices, any tablet PC is really a hybrid of a laptop and a cell phone. Tablet PCs present researchers with interesting innovations in on-site and qualitative research methods. On-site interviewers are actively using tablet PCs to quickly collect (and report on!) data via surveys that may include touch screen components, interactive questions, video or other multimedia components. Gone are the days when surveys have to be programmed and loaded on an actual PC – today tablets can easily access surveys and instantly feed data into online reporting toolsets via a basic wireless connection. Researchers are even using tablet PCs to evolve qualitative research into hybrid

quant/qual techniques. For example, respondents are given a short quant survey to quantify individual preferences, after which survey results can be instantly aggregated and summarized via real-time online reporting tools. Afterward, a focus group discussion of preference or other drivers can take place, incorporating the initial quantitative survey data into the qualitative group discussion.

CHALLENGES As with any new research platform, the cons often outweigh the pros and we’re at an almost dangerous tipping point with social media and mobile. Industry acceptance and adoption is at an all-time high while standards and best practices are still in development.

Respondent privacy is becoming a growing concern as we embrace these new research platforms. Data mining, online behavioral tracking and passive data collection (via device ID) can serve to reduce respondent burden and offer unique opportunities to study ACTUAL respondent behavior (vs. self reported). The question becomes, is this data collected by device IDs and other methods considered “personal” information? Do we need to limit the use of these methods, opt for full disclosure or limit retention of this data? Wandering device IDs, complex privacy policies and fees for SMS/MMS and data streaming present still other challenges.

Also, as market research firms race to be first to market with mobile research services and applications, there appears to be a greater focus on technology than on research quality. In a recent Linked In group, a firm eagerly advertised their ability to handle “long and complex surveys” on mobile devices – a practice we’ve been battling against in online surveys, let alone mobile surveys, as an industry. Respondent engagement on mobile devices is even more challenging than in online surveys which means as an industry we need to be highly sensitive to the mobile survey experience.

Being first to market also presents other challenges. Firms like Thumbspeak and Gongos have both aggressively touted their iPhone survey apps and many more firms are in aggressive development on cross platform applications that will be supported by iPhones, Droids and Blackberrys. Although many of these firms have their own panels to support these apps, the question of “how many survey apps will one person download” looms for our industry. What happens when respondents are inundated with survey apps and a multitude of survey options? As an industry we need to be cognizant of respondents’ thresholds for downloading these apps.

MOBILE SURVEY BEST PRACTICES … MAKING IT WORK Mobile market research presents a challenge to standard online research practices. Studies have shown that in order to get the most out of research efforts on a mobile platform, we need to be sensitive to the unique challenges respondents face when completing surveys on mobile devices.

Small screens, inflexible device-specific user input methods, and potentially slow data transfer speeds all combine to make the survey completion process more difficult than on a typical computer. Couple those hindrances with reduced attention spans and a lower frustration threshold and it's clear that, as researchers, we must be proactive in the design of both the questionnaire and user-interface in order to accommodate mobile respondents and provide them with an excellent survey experience.

Although best practices are still evolving, there are a few that seem to be universally embraced even at this early stage. Here are some best practices to keep in mind when creating mobile-friendly surveys.

Keep the survey short - as in 10 questions or less! It takes longer to navigate on mobile devices due limitations of the user interfaces and data transfer speeds.

 Minimize the number of pages. Each time the page refreshes, the respondent has to wait. This wait time can range from a short to exceptionally long period of time depending on the network connection (with Wi-Fi at the fast end and regular cell data on a poor connection at the slow end). Take care not to put too many questions on a page as mobile devices also have less memory to work with, so a page with too many elements may cause the device to become slow or non-responsive.  Keep the question types simple.  Single dimension radio, checkbox, or select questions are better than multi-dimension grid questions, which could be difficult to complete due to the small screen sizes of mobile devices. Limit scrolling to a single dimension: rows are preferable to columns in order to minimize the need to scroll on both horizontal and vertical axes.  Keep answer lists short to minimize scrolling. Consider putting long answer lists into select (drop-down) questions, but also be aware that these require more "clicks" to complete than radio questions. Select questions only allow a single response, so for multi-select questions, use the checkbox question-type.

 Keep question and answer text short. The small screen and space make it harder to read long options.  Limit usage of open-end questions which require typing.

Minimize all non-essential content It takes extra load time and visual space for every element that appears on the screen. Repetitious elements, innocuous for a typical web-based survey, add an undue burden on respondents completing surveys on the mobile web, in addition to distracting from the purpose at hand: completing the survey.

 Logos: Limit logos to the first (and/or last, if absolutely necessary) page of the survey.  Privacy policies, rules pages, etc.: If possible, also limit these to the first/last survey page.  Progress bar: The progress bar can be very helpful for indicating a respondent's position in the survey, however it also increases the load time and the need for vertical scrolling. Something to consider.

Keep the look of the survey simple. Minimize distraction  Javascript support varies radically for mobile devices, so keep "interactivity" to a minimum. Flash is out for iPhone/iPad/iPod Touch devices.  Some mobile devices also have limited color palettes. High contrast color schemes will work best.

STAYING ON TOP OF THE TECHNOLOGY EVOLUTION (OR IS IT THE TECHNOLOGY REVOLUTION?) Mobile capabilities are changing on a weekly basis. Firms that have already embraced mobile are moving past WAP or app-based surveys and experimenting with DIY software integration, multi-language surveys, geo-tracking, mobile bar coding, digital wallets and image-based survey triggers.

We’ve learned some hard lessons as an industry recently. We need to stay focused on the user-experience. There’s no debate that lengthy surveys are bad for business. Above all, it’s important that as we embrace new research platforms that we build trust and don’t break it. Gone are the days when respondents had no voice. Respondents are powerful and have an incredible amount of influence.

Don’t be left behind! Even if you’re not ready to get into the mobile game, stay in it! Linked In groups like NGMR (Next Gen Market Researchers) and Innovation & New Tools in MR have members that are posting and discussing social media and mobile technologies on an almost daily basis. Industry organizations are including social media and mobile as education tracks in their conferences and for the first time, there is an annual industry conference focused solely on mobile research practices.

In short, embrace cross-platform best practices and stay curious. If you can’t be first in market, be best in market and meet your respondents on the move!

THE AUTHOR Kristin Luck is President, Decipher, United States.

TIME IS OF THE ESSENCE MOBILE SOLUTIONS HOLD THE KEY TO TACTICAL ACTIONABLE RESEARCH Adhil Patel

INTRODUCTION Mobile technology has heralded a new era of research. This new world is not one in which interviewers march around in the blazing heat or freezing cold with clipboards, knocking on doors and surveying housewives for hours on end, trying desperately to extract information about their buying behaviours or their attitude towards the latest new vacuum cleaner. This new world is digital, and this new world is mobile, always on, personal and fast.

We now live in a world where we don’t have to ask people questions to know what they like. We don’t need physical proximity to establish what they’re doing, and with who. And we don’t need an army of interviewers to fuel our ‘data-lust’. Mobile phone penetration is racing towards 100% the world over, with smartphones set to dominate in the near future, so it won’t be long until every person on the planet becomes reachable through digital means.

One side effect of this digitisation of society is the digital shadow that people cast or the digital trail left behind. This information-laden virtual discard pile is where the researcher of the future will need to be comfortable, as time-poor consumers become even less receptive to our invitations to contribute their opinion or capture their purchase and usage behaviour. The savvy researcher of the future will have to be a data detective, specialising in digital forensics and harnessing the power of data mining tools, and tailor-made analyses.

Mobile research is an opportunity. For researchers in developed markets, it offers an opportunity to capitalise on a consumer trend, improve data quality and create better insights. In developing markets, it represents an opportunity to reach respondents that may have been difficult to reach previously, and above all it offers an increase in the speed with which data can be collected, and ultimately analysed. Mobile research is an opportunity for us to make our industry better, and faster.

However, making a success of mobile research does not mean cramming overly long traditional surveys into a new format and continuing as normal. It is not just online research via a new device. Mobile research is cool and it’s quick and it shows respect for a consumer’s time and reflects knowledge of attention spans and the need for interactions to be engaging.

This paper aims to educate and inspire. It aims to discover and deliver. This new area of research encourages researchers to bend and stretch and break the rules. It forces us to modify our expectations, and our perceptions of research. This paper helps the reader to continue the mobile research journey by examining the relatively new area of geolocation research, and taking a look at applications that enable mobile research across smartphone platforms. Tablets are also investigated as a logical extension of the mobile sphere, offering an engaging, intuitive experience.

THE MOBILE RESEARCH LANDSCAPE The term ‘mobile research’ covers a wide variety of types of research. It can refer to a traditional telephonic survey conducted via a mobile phone (Vicente et al, 2009). The more usual reference, however, would be to online surveys delivered to respondents via a mobile device (Ferneyhough et al, 2009). As mobile technology becomes more advanced and pervasive, mobile research is able to adopt new guises. For instance, the inclusion of GPS on smartphones has led to the new area of geolocation or location-based research (Ferneyhough et al, 2010). There are many more though, which makes the term ambiguous and sometimes confusing.

A framework for navigating this landscape is shown in figure 1. This teases out the different ways that researchers can reach respondents (SMS, mobile internet or application), and crosses them with the extent of the respondent interaction (passive, active or social).

FIGURE 1, THE MOBILE RESEARCH LANDSCAPE SMS Mobile Internet Application (App) Passive Click tracking Location and usage tracking, ad exposure, audio sampling Active Survey Survey, panel Diary, survey, polls, scanning, photos, video (ethnography) Social Rate Share Collaborate, compete (Mobile MROC) Patel and Versfeld, 2011

While SMS and mobile internet (including WAP, 3G, etc.) certainly have applicability currently, applications (apps) on smartphones and tablets would appear to be the direction the market is going in. In terms of interactions, active would cover most of the traditional types of research, but passive seems to be an area of white space or blue ocean where mobile research could provide value. These are applications which allow researchers to ‘follow’ respondents without being present, and to collect data in an extremely discreet way, without the respondent having to actually ‘respond’ at all. Passive data collection covers a number of exciting applications, including continuous location tracking (see figure 2), and audio sampling (recording the environment that the mobile is in). Simply put, nothing else does this.

FIGURE 2, CONTINUOUS LOCATION TRACKING USING INTERIOR GPS OR TRIANGULATION WI-FI TO TRACK RESPONDENTS IN RETAIL AND PUBLIC SPACES

Source: Lumi Mobile in Jen and Swinton, 2011

The ability to use the forward facing camera on a smartphone to gauge the reaction of a respondent while they are watching a video streamed to the device borders on creepy, but can be powerful if used correctly. There are already services which allow researchers to identify certain facial expressions that point to particular emotive responses (Jen and Swinton, 2011). Privacy and ethical issues are outside the scope of this paper, but will definitely have a role to play here.

Social interactions with respondents refer to those that are visible to others. An example of this could be a market research online community (MROC) which used a mobile application as the respondent interface, allowing respondents to interact with each other, as well as with the researchers. This would appear to be an easy extension of current MROC research, and is almost certainly being leveraged already.

KEY BENEFITS OF MOBILE RESEARCH As already mentioned, mobile research refers to a wide spectrum of research. Nevertheless, there are a number of benefits of this new area that have been suggested in the literature and by experts in the field. It’s not the aim of this paper to list all the benefits, but rather some of the pertinent ones and particularly those that are unique to mobile research.

First and foremost, the mobile device is personal. While there is still a discrepancy between developed and less developed economies, it will not be long before every person on the planet has at least one mobile device. This device will be in-hand or near the individual all the time, and importantly it’s always switched on. Having a device be unambiguously tied to a specific individual greatly reduces the need for redundancy in surveys, such as demographics and some psychographics, as these will not need to be repeated, but can be linked to newer responses.

What this means in practice is that some hard to reach audiences (geographic or demographic) become reachable, and in an instant. It means that interviewers walking for miles to access rural respondents in Africa and other emerging markets where face-to-face studies are still a reality start to become a rarer sight. While this ability to reach specific audiences can be seen to some extent in developed markets through online panels, mobile research allows for a wider reach than traditional online (because everyone has a mobile, and soon they will all be online), and reduces any possible uncertainty with regards to the target’s identity and suitability (as the device generally won’t be shared, but personal).

Another thing that mobile research can do that is new is provide the opportunity for much richer interactions with respondents. Apart from automatic geo-tagging or geolocation (explained further in this paper), photographs, video capture, bar code and QR code scanning and audio capture all make use of added functionality that can be found in most smartphones today. The fact that most of these will soon boast large touchscreens as well means that whatever the researcher can imagine, in terms of a tactile experience with the respondent, can be programmed into an app. These types of data were not available to researchers traditionally, or at very least were only available at a prohibitive cost.

Added functionality available on mobile devices and the kinds of apps that are now possible also indicate that data quality will improve with mobile research. There are many factors at play, but shorter ‘surveys’, direct digital data capture and increased engagement of respondents will all lead to improvements in overall data quality. Of these, engagement is the only one that likely requires further elaboration. It’s been suggested that just the fact that questions are being asked on the mobile device, which is personal by its nature, increases engagement somewhat. In addition, mobile technologies now open up a new model for research, which is provocatively described as ‘entertainment as research, and research as entertainment’ (Johnson and Swinton, 2011). The idea that research can be part of the actual experience it is measuring goes counter to what many traditional researchers might consider best practise, but it’s exactly this kind of disregard for the ‘rules’ that disruptive innovations tend to display, and is almost certain to engage respondents.

Gamification of surveys has been talked about for longer than most are willing to admit now, without much evidence of game mechanics being implemented on any significant scale (Alberts and Findlay, 2011). This ‘new’ way to capture the attention of the time-poor consumer, and particularly the generation of digital natives with shorter attention spans, can now truly flourish. Conditions are all but perfect for a move by the market research industry to change the old conservative perception that consumers have of us, and to embrace the technology to reposition ourselves as cutting edge and above all relevant.

It’s only a small stretch then to suggest mobile research as a possible channel to be used in customer relationship management (CRM). It would not take much to shape the research instrument such that it could deliver coupons, discounts or special offers. And while it may be a grey area in terms of market research standards, there are surely marketers out there using market research tools and methods (such as MROCs) to push out messages to consumers too. While this may not be an explicit ‘benefit’, it’s certainly a point to consider, and to be aware that others may be offering it.

Traditional research methods involve an interruption or appointment method (Friedman and Smith, 2010), and involve co- ordinating two (or more) parties to be ‘present’ or at least participate. Mobile research overturns this idea, enabling instead a kind of solitaire or solo research experience, which can be started and stopped at the convenience of the respondent. This allows for much more efficient and less intrusive use of time, as respondents can choose to participate during what can usually be considered dead- or down-time. This might be on a commute to or from work, during a lunch hour or break, or indeed whenever the respondent chooses. The flip side of that coin is that the time and place for responding can also be stipulated by the researcher, if necessary. For example, gauging the crowd at a music festival or sporting event could be done in real-time, using geolocation to verify attendance. This real-time nature of mobile can reduce survey-lag, which refers to the length of time that elapses from the event a respondent is expected to recall (shopping trip, purchase event, specific television program, etc.) until the survey is conducted. This lag may well be of significant length, increasing the chance of memory failure, and memory reconstruction. Memory has been shown to be rather malleable and unreliable (Braun-Latour and Zaltman, 2006), so reducing this lag will bolster data quality further as well.

Lastly, it’s worth mentioning speed. For many, this is the primary benefit of mobile research, while for others it is not seen as a benefit at all. While there can be little doubt that mobile research approaches should be fast, the question remains whether they are significantly faster than current methods. This of course will depend on what the comparative set is. Comparing to face-to-face interviewing in rural Africa will be quite different to comparing to online panel research in Germany. One way or the other, the speed of mobile research plays a definite role, but it may or may not end up being a unique benefit.

EXPERT PANEL In an attempt to obtain a more considered and perhaps balanced view of the potential of mobile research, a small expert panel was put together of eight research practitioners with significant mobile experience. A mix of prominent industry voices (Ray Poynter, Guy Rolfe and Lenny Murphy) were joined by suppliers of specific mobile services (Rolfe Swinton of Lumi Mobile, George Wigley of Qriously, and Emmanuel Bellity of HeyCrowd). Rounding off the panel were James Fergusson (Head of Technology) and Tyson McKeown (Africa Middle East Operations Director), both from TNS.

On the whole, the panel seems to agree that mobile is moving towards the main stream relatively fast, with the majority suggesting that it will overtake traditional online research in two to four years. Given how small it is currently (at about 1%) it’s got a long way to go (Johnson and Swinton, 2011). There was also largely agreement that geolocation or location- based mobile research is likely to have the biggest impact on market research as a whole, beating out ‘mobile as an interface for MROCs’ with none voting for ‘two-way video’ (recording of responses to video via mobile device). The panel was divided, however, on the biggest benefit of mobile research. In the end, speed was the clear winner, with data quality and the ability to reach niche audiences splitting the remaining votes.

FIGURE 3, WORD CLOUD OF EXPERT PANEL RESPONSES REGARDING THE FUTURE OF MOBILE RESEARCH

As shown in figure 3, there were various views on where mobile might have an impact in the near future. Key themes observed were questionnaire length, inevitability and speed. In some ways, mobile research could be the methodology that finally forces researchers to shorten surveys to a length that is manageable for respondents. However, mobile research of the future will not be traditional surveys shoe-horned to fit on a mobile device. Surveys will be custom made, with proper considerations made for the platform being used (Ferneyhough et al, 2009). The certainty with which the panel spoke of this mobile future hinted at a realisation that the point of no return is behind us. This was summed up fantastically by one of the panellists: “Mobile is the future of market research – period! Either accept it and learn to grapple with it or be buried by it!”

TAKING A CLOSER LOOK AT SPEED It’s worth exploring the speed of mobile research in a bit more depth. As mentioned earlier, it may appear to those in developed markets that there are no real gains when moving from traditional online research to mobile research. However, in many developing markets, the relevant comparison would actually be from face-to-face research to mobile survey research. In this case there would be significant gains.

The trend amongst mobile research providers is to have an end-to-end system, incorporating an online user interface through which questions can be deployed, and data received, often in chart format (see figure 4). The end-to-end system reduces the number of human interactions needed in the research process, by removing scripting, printing and data capture (where relevant), and in many cases data processing and charting too. Thus, the system will deliver improved data quality, but also a faster, fully digital process. Once the recruitment is included into the system, which is easily done with mobile, the process becomes truly seamless.

FIGURE 4, ONLINE USER INTERFACES ARE BECOMING MORE COMMONPLACE FOR MOBILE TOOLS

While survey research and other approaches with traditional equivalents (like diaries) are easy to compare, other facets of mobile research are not as clear. It doesn’t really make sense to talk about the speed of audio sampling, or geolocation research in comparison to traditional research, as these do not have natural equivalents. When it comes to photo or video tasks (broadly ethnographic), these studies would have required dedicated hardware, briefings, much planning and generally large incentives. It would seem that mobile research offers large gains in speed here, as well as cost (given respondents will use their own devices).

The implication, or end result of having access to faster research is that clients can finally make tactical decisions based on the information gleaned from it. Trackers, for example, are notoriously slow, with data often only reaching clients months after the interviews were conducted, and generally quite long after they would have been useful (Patel, 2010). With mobile research, the opportunity now exists for research to be more usable, and actionable, as results can be delivered almost in real-time, reducing the time from deciding to ask certain questions until receiving answers to a more acceptable level.

PAY AS YOU GO BUSINESS MODEL One area that was only mentioned in passing by one of the panel members was that of cost. This may have been an attempt to stay away from a potentially contentious issue (money always is, isn’t it?). It seems more likely that the panel had all but decided that mobile research would not offer much of a cost discount. Clients may scoff at this idea, possibly suggesting that market research companies are pocketing the gains, but it would appear that going mobile will not be reducing costs to client – at least not yet, or as a panel member suggested, not ‘as currently offered’. One may be tempted to believe that the big market research companies, perceived to be somewhat slower than boutiques to take up innovations, would be slow to start offering mobile research. However, most of the big companies are already playing in the arena, largely on the back of partnerships with relatively small mobile research service providers. This partnering model works so well in mobile that even these providers sometimes partner with niche providers of technical solutions (Jen and Swinton, 2011). In this way, researchers have access to skills and technology that would not have been feasible in-house, at least in the foreseeable future.

Nevertheless, there are a number of very focussed mobile research providers that don’t have the overheads of large corporates and have specialised expertise in the mobile area. These providers can and do sell directly to clients. Indeed, clients do not even need to ever deal with the (human) provider at all, in many cases. User interfaces are taking the place of research executives, allowing clients access to a DIY (do-it-yourself) research paradigm, in which all extra costs are cut out of the system. At one level this is unfortunate for the research firms, as they miss an opportunity to provide insights (and a claim to the revenue), but it also puts additional pressure on the client that may not have the headspace or bandwidth to spend on analysing the data themselves.

That aside, the pay-as-you-go business model is one that seems to fit well with mobile research involving survey questions, but other types of mobile research may have more complicated costing structures. Charging a per-response fee

makes the costing quite neat and determinable up front. Costing per question also tends to keep the questionnaire length down, as there is incentive to keep the number of questions to a minimum. This more transparent approach to costs will surely appeal to procurement departments, but has the potential to turn this kind of research into a commodity once there is enough competition.

AREAS OF INTEREST Mobile research is still relatively young, with widespread penetration of mobile devices only occurring within the last ten years. It’s hardly surprising then that the literature on the topic is rather sparse, and the papers that do exist tend to deal with relatively basic ideas and applications (Ferneyhough et al, 2009; Bellity, 2010; Fisher et al, 2009). The pace with which technology moves, however, tends to outstrip the rate at which papers are published, and hence many new applications show up as services being offered to researchers, often by small focussed companies, before they are presented at conferences.

Three specific areas of interest are covered here, which seem sensible uses of mobile research.

GEOLOCATION RESEARCH Location-based mobile research, sometimes just called geolocation research, may well be one of the new methodologies that mobile research brings to market research, which could eventually capture a significant proportion industry. Given the rise of smartphones, and the subsequent prevalence of GPS, it’s now possible to identify the (almost) exact location of any given respondent (or at least their handset) at a given time (Ferneyhough et al, 2010). Tying survey-type information to the location will, when GPS becomes accurate enough, allow marketers to be much more precise in their targeting.

While the idea of tagging the location of a respondent to their data is not a new idea, the speed with which services can now produce these mapped deliverables is notable. As previously mentioned, this speed can usually be attributed to the end-to-end nature of the process, and the automation of the majority of the research process. It’s not uncommon now to be able to capture responses to questions in minutes, watching the deliverables populate themselves in real time. The speed of this type of mobile research can lead to actions being taken by marketers (tactical) that were not possible with more traditional methods. There are almost certainly applications for this kind of research in sponsorship effectiveness and event evaluation studies, where location can be important, in addition to a number of other scenarios as described by Ferneyhough et al (2010). The ability to reach the respondent during or soon after an experience will ensure a fresh response to questions, without having provided an opportunity for memories to falter, as discussed.

FIGURE 5, REAL TIME DELIVERY OF GEOLOCATION DATA “DO YOU CARE ABOUT PRIVACY ON MOBILE PHONES?”

Qriously

A pilot was conducted, to test this methodology, making use of a new service called Qriously. This particular service inserts questions into applications (e.g. a photo application), rather than making use of a dedicated market research application. While the service is still in pre-release, response rates are reported to be higher than banner ad click rates,

though the questions still need to be served to a large number to achieve a reasonable sample. The make-up of the sample is thus dependent on the reach of the applications in which the questions are being placed.

FIGURE 6, QRIOUSLY: IN APP QUESTIONING LINKED TO GEOLOCATION AND REAL TIME REPORTING

The pay-as-you-go business model fits well with this type of research, removing any barriers there might be to really quick research. The costs also tend to be quite low, paying a small fee per response, making it possible to use this research to complement more traditional work, without adding significant cost or time.

Given that most mobile apps now have the ability to capture the GPS location of respondents, it’s possible for the researcher to link this data to a map outside of the research application, if that was necessary. There are some easy (and free) websites that convert the data to the format required for adding to Google maps, for instance.

For the purposes of this paper, research experiments were conducted using various research services. To create a link between these experiments, it was decided that certain questions would be repeated. Questions relating to influence, specifically looking at friend networks on Facebook, were asked. See figure 7 for results from the in-app, single question format, through the Qriously platform.

FIGURE 7, LARGE SAMPLE (N=1000) OBTAINED IN JUST OVER AN HOUR, FROM QUESTION CREATION TO CHART: HOW MANY FACEBOOK FRIENDS DO YOU HAVE?”

While multiple question formats were available, the slider seemed most appropriate for this question, and anchors of ‘0’ (on the left) and ‘600+’ (on the right) were used. A sample of 1000 was collected, from across the globe, in an extremely short time (75 minutes).

Although the interpretation of the results was not the point of the pilot, it was interesting to note that the average was slightly to the right of centre, suggesting an average above 300. The control samples both gave similar averages, at 352 for the more standard mobile survey (using the HeyCrowd app), and 313 for the traditional online survey. HeyCrowd also allowed for capture of geo-location data, and therefore mapping of these responses too, though not automatically at this time.

Gut feel: Exciting application, need to establish the right research needs.

CROSS-PLATFORM APPLICATIONS The difficulties inherent in achieving a consistent survey experience have been examined in previous papers (Fisher et al, 2009; Ferneyhough et al, 2009). However, it is difficult to find mention of applications that cross platforms, or even just operating systems, enabling delivery of surveys to a large variety of phones. Indeed, most services currently operate on only one operating system, with development being done to extend capability. Being able to reach respondents irrespective of whether they use an iPhone, a Blackberry or an Android handset is surely important if mobile research is to be taken seriously. However, this is not a trivial task. Just examining, for example, the variance in experience of a consumer application like Facebook, one can immediately see that achieving a consistent experience will be a difficult endeavour.

In developing markets, where smartphone penetration is currently lower than elsewhere, cross-platform may refer to the ability to serve a similar experience on a feature phone (a more basic handset). Often, these phones will have mobile internet capabilities (WAP or GPRS) without the user making use of it, or even knowing much about it (TNS, 2011). Some services have managed to deploy applications across platforms effectively, but this does require significant investment in terms development (Johnson and Swinton, 2011).

Presently, it would appear that conducting a cross-platform study (vs. single-platform) increases costs dramatically and usually requires extra time too. Research houses wanting to deploy this kind of service need to invest much time and effort for training and harmonising the system with their existing systems, not to mention the costs of licensing.

One might expect that the end-to-end ability including pay-as-you-go costing would make sense in this situation. However, for many large research companies, this arrangement would not afford enough control. The compromise may well be a

customisation (and co-creation) of the system, affording all the checks and control needed by the research house, but also the efficiency and nimbleness of the mobile service.

Gut feel: This ability is currently rare, but it’s critical for mobile research to be able to go mainstream.

TABLETS These cross-platform considerations are further complicated by issues relating to form (e.g. handset vs. tablet). The tablet is still quite early in its adoption cycle, with the Apple iPad (into its second manifestation already) still dominating, though Android tablets and a variety of others have now made this quite a competitive market. The fact that these tablets have much larger screens than even the largest smartphones means that applications need to be made specifically for the tablet to optimise the experience.

Operating systems tend to be common for the tablets and smartphones (e.g. iOS is used for iPhone and iPad), which makes the ecosystem somewhat less complicated than it could have been, but this also means that applications are sometimes shared between devices. Stretching an application to fit a tablet screen tends to distort the interface, and a resolution mismatch occurs, making icons pixelate and images (including video) appear fuzzy. The alternative is to view the app at handset size, which defeats the purpose entirely.

The literature is quite sparse on tablets, and particularly so for market research conducted on tablets. This is not surprising, given that there are only a handful of purpose built survey applications that try to optimise for the tablet. Of those that do exist, most are focussed on fieldwork being conducted via interviewers capturing responses on tablets, rather than the respondent completing the survey directly. Again, this fits with the low penetration of this type of device globally (TNS, 2011). In addition, these applications tend to be drab and unexciting. Aesthetic improvements would greatly increase their appeal.

The downward pressure on tablet prices and the sheer amount of competition will lead to an explosion of the tablet market over the next few years, putting these devices firmly at the centre of the digital universe, alongside the mobile phone. This means that research will have to adjust to become more relevant when competing with other entertaining alternatives, all vying for the attention of the potential respondent in a multi-screen environment (Johnson and Swinton, 2011). Thinking of the tablet as ‘just a bigger mobile’ will almost certainly lead to failure.

Gut feel: This area has much potential, and is crying out for a gamified and optimised research experience.

AREAS FOR FUTURE RESEARCH The passive end of the mobile spectrum does seem to have some of the most exciting applications. Being able to ‘follow’ respondents around, see how they’re using their mobiles, ‘watch’ their reactions to experiences, events and marketing activities all feel a little bit like science fiction. Listening in (audio sampling) on consumers or giving them tasks to complete in real-time, on the other hand, brings a ‘Big Brother’ element into the mix, and opens up the debate on privacy.

Today, innovations happen so fast, with only slightly slower adoption rates. The flip side of this is that some approaches may well lose relevance as fast as they were introduced, and it’s difficult, if not impossible to predict early on which ones will succeed, and which will fail. This makes it unwise for researchers to make massive bets on an uncertain future, suggesting instead that many small experiments would be safer in the long run. Having said that, researchers should be ready to jump on trends (not fads) that have managed to take hold, and to this end should be keeping their ears to the ground, and listening to their clients too.

With this philosophy in mind, innovators and early adopters in mobile market research should be looking to establish where mobile can carve out a decent sized niche for itself, rather than going head to head with traditional techniques, at this stage. Geolocation has been mentioned by many, and further research into this approach should attempt to uncover further benefits to ensure easy uptake by the industry. How it can better complement traditional surveys, and how (or if?) it can stand on its own are both questions that further research could answer.

Another extremely interesting application is in biological or medical research, where the device can be linked to sensors on a person’s body. While this has many apparent benefits for patients under observation, and for health in general, when these sensors become more mainstream, they can also be used to obtain more holistic measurement of responses to experiences, brands etc. In essence, mobile could be a way to make neuroscience (at least some form of it) more portable, and more accessible to market researchers.

Lastly, it seems a reasonable extension of mobile research that the data being collected through these mobile devices need to be analysed. At present, our industry makes use of a tiny proportion of the data that could be available, and this information glut is only going to deepen for the foreseeable future. As an industry we are short on analysts that are truly able to mine insight from these streams of data, and our techniques are inadequate for dealing with such large datasets, and complex structures. Being able to make sense of the flood that will be unleashed by mobile research will be crucial. Big data is definitely an area that warrants some further research, even if just to establish whether or not market research is the right home for it.

SUMMARY Mobile research is an exciting (relatively) new area, which is set to make a big splash in the market research world in the short to medium term. However, there are various different facets to this diverse research area, each with its own pros and cons, and with its own pace of adoption. Surveys that are similar to traditional online surveys can easily be adapted for mobile, and this facet is probably the most advanced, followed by applications built to act as consumption or usage diaries. Geolocation has further to go to reach the mainstream, though it’s arguably moving at a faster rate. In terms of format, crossing platforms continues to be an ability that is offered by few, though is crucial for the adoption of mobile. Research using tablets is starting to emerge, but there is a long way to go, and proper optimisation is still elusive. In the end, above all, mobile research is exciting, fast and an opportunity for our industry to transform itself and capture the imagination of a new generation. Let’s go out and do it, together!

REFERENCES Alberts, K. and Findlay, F. (2011). Gamification: What It Is, and What It’s Not. Paper to be presented at ESOMAR Congress, Amsterdam, Netherlands. Bellity, E. (2010). Innovative Mobile Research in Developing Countries. Paper presented at ESOMAR: Online Research, Berlin, Germany. Braun-Latour, K. A. and Zaltman, G. (2006). Memory Change: An Intimate Measure of Persuasion. Journal of Advertising Research, March 2006. Ferneyhough, C., Grenville, A. and Morden, M. (2010). Mobile Location Based Research. Paper presented at ESOMAR: Online Research, Berlin, Germany. Ferneyhough, C., Ryan, C. and Zahariev, M. (2009). Best Practices in Mobile Research. Paper presented at ESOMAR: Online Research, Chicago, USA. Fisher, S., Jellison, C., Lavine, S.R. and Thyagarajan, U. (2009). Mobile Interviewing: The next frontier of data collection. Paper presented at ESOMAR: Online Research, Chicago, USA. Friedman, L. and Smith, M. (2010). The New Influence Tracking. Paper presented at the 56th ARF Convention, New York, USA. Gichuri, J. (2011). Mobile & Digital Technology & the Application of Needs-Based Research in sub-Saharan Africa. Paper presented at the 32nd SAMRA Convention, Vanderbijlpark, South Africa. Jen, J. and Swinton, R. (2011). Leveraging Digital Insights via Mobile Channels. Paper presented at the 2nd International Conference on Market Research in the Mobile World, Atlanta, USA. Johnson, A.J. and Swinton, R. (2011). Developing Second Generation Mobile Research Techniques. Paper to be presented at ESOMAR Congress, Amsterdam, Netherlands. Joubert, T., (2006). Venturing beyond the boundaries of conventional sampling techniques… and beyond the borders of South Africa. Paper presented the 27th SAMRA Convention, Dikhololo Game Reserve, South Africa. Patel, A. (2010). The Future of Brand Tracking Must be Leaner and More Focused: A Radical New Approach. Paper presented at ESOMAR Congress, Athens, Greece. Patel, A. and Versfeld, R. (2011). Mobile Research: iFuture or iFlop? Paper presented at the 32nd SAMRA Convention, Vanderbijlpark, South Africa. TNS. (2011). Mobile Life: Global Telecomms Insights. (discovermobilelife.com). Vicente, P. Reis, E. and Santos, M. (2009). Using mobile phones for survey research. International Journal of Market Research Vol. 51 Issue 5.

THE AUTHOR Adhil Patel is Head of Thought Leadership, TNS Global Brand Equity Centre, South Africa.

THE PLACE FOR MOBILE RESEARCH? MULTI-MODE STUDIES OF MAJOR CULTURAL EVENTS Simon Atkinson • Sean Conry

INTRODUCTION For a brief moment, the wedding of Prince William and Kate Middleton on 29th April was the news event. It attracted a worldwide TV audience of more than two billion people, including 72 million viewers streaming it on YouTube, and many worldwide newspapers and magazines giving unprecedented profile to the event. The experience in the UK – where the public was granted an official holiday - was of course more intense. In addition, tens of thousands of people flocked to London to catch a glimpse of the event, while approximately 30 million people watched the event on TV, the highest figures since the funeral of Princess Diana in 1997.

Enter the researchers. We are trained to take “pre” and “post” measurements, to evaluate an event like this. But mobile apps now enable us to measure “the moment” as well. With the tipping point in smartphone ownership now in sight, the authors believed that mobile research could provide an important and complementary channel.

Mobile research is still an emerging field, but to date much of its application has been to assess shorter point-in-time events and occurrences, such as point-of-purchase. But events like this have as a fairly long duration (i.e. a whole weekend), in addition to significant lead-up and denouement. Attitudes and opinions change throughout each of these stages, including the day of the event itself.

Thus, the authors believed that to fully understand the effects of a major cultural event such as this, that multiple channels of data collection had to be used. Simply put, no single method could give us a rounded sense of what people thought about the Wedding.

An opinion poll, for example, can give us insights into underlying values. It can track attitudes over time and it can give us instant “opinions” on the stories of the day. But surveys like these do not get us close to the respondent – for example to find out more about how people really feel, and how they behave.

Conventional focus groups are of course one way to start to get deeper. In the case of an event like The Royal Wedding, we could organise a focus group in advance and ask people to project ahead to the weekend. Or we might run some groups reasonably soon afterwards, and elicit post-hoc assessments. Both have their merits, but for this event, we wanted to get a little closer.

In particular, we wanted to explore what happened once the day got going, and the coverage started in earnest. How did people feel at particular points of the weekend? Happy? Sad? Bored? Some in the media reported polls and voxpops saying “people weren’t interested”. Were they right?

Traditionally we might have measured sentiment before and after the event, to help us test some hypotheses, but the emergence of mobile digital qualitative research now gives researchers the unprecedented ability to see how attitudes, opinions and sentiment evolve throughout the duration of an event.

This paper shows how we took advantage of the latest in survey technology. Through the use of a research app, we set our respondents the task of reporting back on what they were seeing and how they were feeling over a three day period. This method gave us the opportunity to collect data of many different types, both qualitative and quantitative, and obtain in- the-moment perspectives.

But we had questions. Could we meaningfully aggregate all the data? How could we present it at the speed required for newsworthiness? Would we learn anything more by combining all these modes or would “keeping it simple” have been better?

THE RESEARCH APPROACH

Summary of the research methods Goal Method

Collect baseline of public attitudes Traditional opinion poll (telephone)

Measure what was being said and where before, during Online listening and after the event

Understand: A) the citizen experience of the event itself; Provide respondents with a mobile app and set them the and B) how perceptions and attitudes change throughout task of sharing their experiences in real time the course of a two to three day event

Data Analysis and Reporting  Identify key themes to discuss in Sky News TV interview  Collate qualitative and quantitative information from the mobile research  Incorporate and link GIS data with other research findings from polling and web listening for integration into a meaningful report

1. Opinion Poll In the run-up to the event, we conducted a classic opinion poll for Reuters which gave us the baseline of where public attitudes to the monarchy stood in April 2011. Since Ipsos MORI had been running some of the questions for 20 years, we were able to see how attitudes had changed over time.

This survey is grounded in the principles of good quantitative research, first developed in the 1930s and 1940s.

 We interview, at random, a representative sample of people across Great Britain.  We ask them a series of carefully constructed questions.  We include questions that had been asked in previous surveys, thereby putting the results in context.

The intention is to have a robust and representative picture of what the British public are feeling, at a particular point in time.

For example, we find that anyone coming up with assertions or headlines along the lines of “the British public have had enough of the monarchy” are wide of the mark – and have been for many, many years (see figure 1).

We also find the public divided in their views as to whether Prince Charles should give up his right to the throne in favour of Prince William (see figure 2).

But we also noted, ahead of The Royal Wedding, certain apathy toward the event itself, something which was particularly evident among younger people. This was picked up widely by the media. (See figure 3.)

FIGURE 1 FIGURE 2 BRITAIN SHOULD REMAIN A MONARCHY? SHOULD CHARLES GIVE UP IN FAVOUR OF IT’S BEEN LIKE THIS FOR TWENTY YEARS! WILLIAM? IT EBBS AND FLOWS …

FIGURE 3, 12 DAYS OUT HALF SAID THEY WERE INTERESTED

2. Online Listening There are some parallels here with elections, where pollsters grapple with voting intention surveys conducted some days before polling day. Our question was this: would this less-than-excited picture be replicated on the day, or will a different pattern emerge as the event unfolds?

One clue here was in what we heard when we “listened in” on-line. We learned that interest was increasing. The concept of the “Street Party” became more and more talked about, as local authorities made special arrangements and organising committees got going. (See figure 4.)

FIGURE 4, MENTIONS OF “ROYAL WEDDING STREET PARTY”

Having established that purchase decisions are emotionally driven and that much of our shopping is habitual, The Marketing Store believes that there is a great opportunity for retailers and brands to develop stores and promotional communications that are better able to connect emotionally with their customers. They therefore approached BrainJuicer to run a pilot study to understand the emotional context of the shopper as they went about their shop. Specifically, through this piece of work, they set out to establish:

We measured what was being said, by whom and on which sites. This covered news, blogs, forums, video, Twitter and other sites. In the UK, mentions on Twitter represented 39% of the 771,499 mentions counted. This “social listening”

painted a vivid picture of how activity built up over time, peaked, and then died away. It puts our opinion poll in some further context. Our fieldwork dates of 15-17 April can be firmly placed in the relatively quiet “pre-Wedding” period”.

Figure 5 illustrates mentions simply of the term ‘Royal Wedding’. This predictably increased in the days prior to the event, peaked on the day and fell away after the first Monday when the nation was back at work.

FIGURE 5, “ROYAL WEDDING” ONLINE MENTIONS: SHARP PEAK ON 29TH APRIL …

Data is provided courtesy of Brandwatch Social Media Monitoring.

This poses the question, of course, as to whether we could get closer to what was happening during this peak in activity over the weekend itself. Of critical importance is the potential for opinions to change en masse in such a focussed time frame, and so we wanted to know if we could get closer to the experience, thoughts and attitudes of the public. To do this, we needed to have a “bottom-up” approach, and get ourselves connected first-hand to what people were doing.

3. Mobile Applications Mobile apps offer researchers the ability to capture “in the moment” stories, changing sentiment and trends in experiences of larger groups. Two things helped us build on the findings of the opinion poll and the social listening.

First, the increasing prevalence of smartphones, now owned by around 30% of the UK population, meant that we could reach a large audience of people who were away from their computers or fixed line phones.

We know that people report using their smartphones between one and 10 times an hour (see figure 6,). So we knew there would be potential for applications as a powerful channel for accessing public opinion. For example, people who use Facebook on their mobile devices are twice as active as non-mobile users. There is a need for using their mobile devices to provide updates on their experiences in real time. We wanted to emulate this method of communication for our project over the Royal Wedding weekend.

FIGURE 6

Getting the right balance was important. We took a “middle line” approach in our prompts and communications in order to balance maximize compliance without being overly intrusive.

We prompted them at key points during the weekend (for example reminding them on the morning of the Wedding) but did not communicate too frequently. One of the key considerations for researchers implementing this method will be the extent to which they want to “accompany” people through a process such as this.

One option is the simple prompting of respondents with generic encouragements and reminders (for example via SMS or on-device triggers, alarms and notifications) to get them to complete tasks at the appropriate times. Another possibility is to respond to individuals’ contributions with more tailored prompts – for example to particular groups (who can be asked more detailed questions), or in response to people who say something particularly interesting (who can be asked follow-up questions). Thus, there is a spectrum ranging from broad and generic to quite specific and individualized prompts.

Recruitment We knew that the App was not something that needed to be restricted to a small group of, say, 15-20 people. At the same time, recruiting, from scratch, a group of hundreds (or thousands) of respondents to help us would have been prohibitively expensive. We needed to find a balance between statistically relevant information and a manageable amount of data that would be efficient and valuable. The app was targeted to run on iPhone, Android, Nokia, Blackberry and Windows Mobile devices in order to allow for as many people as possible to participate.

An existing Ipsos panel consisting of more than 400,000 members in the UK who we approach for much of our quantitative research work was the source for our pool of mobile participants. They had previously been pre-screened and asked about their device ownership as well as their technology usage. Thus, we were able to send out a targeted invitation. The email was sent to around 15,000 people and they were asked to complete a short questionnaire with some further questions about the specific handsets they were using.

Our recruitment process yielded some 754 panel members who said they would take part. They formed our base sample for the research over the weekend.

Research Design The app gave us the opportunity to gain information from the respondents in a range of different formats, and so we built these into the research design. The feedback mechanisms included:

 A structured survey which was completed after the Wedding. We could then compare these results with the findings of the pre-survey and perform a classic “pre versus post” measurement.  Short “point in time” surveys to capture how people were feeling or what they were doing – participants could complete these as many times as they liked. For example, they were invited to tell us how they were feeling: bored, embarrassed, excited, impressed, sad, etc.  A form to complete their “Royal Wedding moments” – enabling us to collect more qualitative data. Again, respondents could submit as many times as they liked. These were similar to text messages.  The ability to capture photos of what was happening and send them back to us. These photos were then uploaded onto the website of our media partner at Sky News, creating a form of citizen journalism, and forming the basis of a TV interview on the evening of Friday 29th.  An opportunity for respondents to (with their consent) share their location with us via GPS information, thereby enabling us to track the location of our respondents over the weekend. (See figure 7.)

FIGURE 7, RESPONDENT EXPERIENCE

OBSERVATIONS Predictably there was a peak in feedback received on the day itself.

Activity peaked in the immediate run-up to the Wedding in the moments of anticipation between 10 and 11am when guests were arriving at Westminster Abbey. When the event itself was happening, numbers dropped away - presumably because respondents were either experiencing it, talking with the people they were with or engaging with their usual social media channels to discuss what was happening. We had consistent activity for the following six hours, as we kept in touch with people through the weekend – for example reminding them to send us their “Royal Wedding moments” and later in the weekend, to complete the post-wedding questionnaire. But we didn’t “push them” to continue participating after Saturday lunchtime. It was clear that the event was certainly over by Monday morning, at least in the eyes of the public! (See figure 8.)

FIGURE 8, ACTIVITY: PEAKS AND TROUGHS

Perhaps the strangest finding from the qualitative dimension was the way people got drawn into the event. People went from:

“I’ve no plans as such … waiting to see the weather. If I’m in I’ll have the wedding on in the background. Just to see Kate’s dress. I’m avoiding the royal wedding at all costs!”

To:

“I’m surprised. I started crying when I saw the two princes!”

“I’m a bloke but suddenly struck with excitement, now that Kate is almost there. I really didn’t think I’d be watching but the majesty of it all is so absorbing”

Perhaps the bonds of family helped. One of the things the research app provides us with is real-time information on what people are doing. Or, as in the case below, who they are with (see figure 9).

FIGURE 9, REAL TIME INFORMATION! PLEASE SPECIFY THE OTHER PEOPLE YOU ARE WITH …

Privacy We noticed that, once people shared their location with us via GPS, they shared it multiple times, indicating a broad cell of trust given participants were drawn from our research panel, we were perhaps in a better position than if we had recruited a group of “first-timers”. Of 783 wedding moments voluntarily submitted, there were 505 GPS readings. Still others completely avoided capturing their location for us. Most people were enthusiastic about sharing photographs, sometimes of their television, other times of their friends, family or themselves. We gathered 218 photographs within the short survey which included questions about how they felt right now and who they were with. Combining the photo with the qualitative photographic data helped to more easily code the content of the photographs.

People gave permission for their photographs to be used publicly and so those submitted were immediately forwarded to Sky News who uploaded them to their website, painting a picture of how British people were experiencing the Royal Wedding. The immediacy of this data was very powerful and was a strong talking point at Sky News that evening, culminating in an interview with Ipsos MORI’s Sarah Castell. Rather than online listening, these photographs were gathered alongside structured survey questions to give a coded, validated perspective in real time.

Geolocation data is a relatively new option in the researcher’s toolkit. On the aggregate, it helps to illustrate the geographic balance of a research sample. For example, the map illustrated in figure 10 shows the wide range of locations where participants were to be found throughout the day:

FIGURE 10

In addition, geolocation data also has power at the individual level; it can be utilised to tell a rich, geospatial story of the respondent experience. While latitude and longitude coordinates are described in numeric integers, we prefer to think of GPS data as qualitative data. It cannot be cross-tabbed to reveal significance, but it can help tell a powerful story.

Our first illustration follows a user’s “pre-wedding” mobile questions, through to a collection of their wedding day moments, and concludes with the mobile Post-wedding mobile input. This participant in particular shows how emotions, attitudes and opinions can shift throughout the course of events of major cultural significance, even over a period of time as short as a day.

Initially, our participant appears to have no plans or expectations for the day. However, the decision to attend a community gathering and “get involved” resulted in an expression of national pride. (See figure 11.)

FIGURE 11

Another example shows people who planned their day around community interaction and making an event of the day. (See figure 12.)

Whether seeking out like-minded individuals, or ambivalently joining a gathering, a common thread of the sense of importance of community emerges, and we can see how creating a destination plays a part in the collective experience amongst the population of such a major cultural event.

As we start to look at these patterns in more detail, the importance of having people explicitly opt-in to such an exercise cannot be overstated. While the lines are still blurred about exactly what constitutes public or “personal information” that can be used for research purposes (for example with opinions and photographs uploaded to open social networking sites), this method of collecting data allows researchers to be certain that the content generated can be used for market research purposes.

Moreover, the combination of self-initiated reports with directed surveys allows researchers to: a) capture the unclouded respondent experience from their perspective, while (b) keeping open the possibility of issuing direct tasks and surveys to help answer the specific business questions that need answers. (See figure 13.)

FIGURE 12

FIGURE 13

Timing Something which comes through very clearly – and which would be good to test in other contexts – is in the peaks and troughs of how people feel during an event such as this.

It also appears that, the more people feel something (positive or negative), the more they are likely to share it – as we see by the number of responses we started to get back describing how they were feeling. This is something which many of us will have seen in our own lives through Twitter and Facebook, and it appears to translate to researching people using the approach we adopted here. (See figure 14.)

FIGURE 14, A PEAK IN FEELING EXCITED – BUT HAPPINESS AND PRIDE WERE MORE SUSTAINED, ESPECIALLY AFTER BALCONY SCENE

Findings Media coverage ahead of the event ranged from the excitable (“a nation awaits”) to the gloomy (“no one is interested in the Wedding”). We were looking for, and found insights on, a few key questions:

1. Are the British public less pro-monarchy than they were (answer: No); 2. Were they disinterested in events (answer: They pretended not to be, or didn’t think they would be, but got drawn in); 3. Did they feel a sense of pride (answer: Yes, on balance).

No single research method would have sufficed in this project, because people experienced different emotions and thoughts over time, from the run-up to the Wedding, to planning how they would spend the day, to how individual moments throughout the day impacted their mood and finally, how the experience affected them the day after – with implications for long term sentiment toward the monarchy.

We have some compelling evidence that people enjoyed the Wedding more than they had anticipated (see figure 15).

Since responses tailed away very quickly after Saturday lunchtime, it’s not clear whether, had we not conducted this research, we would have missed quantifiable evidence of what people were telling us at the time. We would instead have had to rely on post-wedding surveys and/or anecdotal evidence and/or what journalists decreed to be the case.

By carrying out a multi-mode study we were able to cover the build-up, peak and aftermath, and reach people at different times throughout the course of their response to the event.

One final observation on our respondents: we found, through comparing our pre-wedding survey with the findings of the opinion poll, that our group was more pro-monarchy going into the weekend than the public at large. And the event itself only served to make them even more convinced that the monarchy is here to stay (see figure 16).

FIGURE 15

FIGURE 16, WILL BRITAIN HAVE A MONARCHY IN 50 YEARS’ TIME? MORE CONVINCED POST-WEDDING!

LESSONS FOR RESEARCHERS AT LARGE We believe there are some useful observations to report on the lessons learned from the experience, particularly in relation to the use of the mobile app in context of a mixed-mode research project. Using an app to complement an opinion poll can certainly be applied elsewhere, for example in elections, awards ceremonies and sports matches. It can also be used in longitudinal work over time – for example in understanding the experiences of particular segments such as young people.

And there are potential uses in helping better understand the innovation process (testing out people’s reaction to a new service or product), or day-to-day experience (to better understand the customer journey, for example).

Here are some questions we have been asking ourselves:

1. Can we engage “regular respondents” in mobile research? The answer is a qualified “yes”. Our participants were drawn from across the country, including 25% from working class backgrounds. However, the response rate, even from an access panel of people who are comfortable with the idea of taking part in surveys, is not particularly high. The need to really “sell” what you are doing, and why is paramount. It is also critical to keep a conversation going with respondents. In some instances, this might be described as a “dialogue”. In others, it is likely to be a matter of sending out reminders in order to achieve the required sample.

2. Is this interesting? We found that once people had downloaded the app and got going, they took part on multiple occasions. For example, we received 1,297 “surveys” from 287 people. The key issue is translating the initial interest into downloading the app and getting people comfortable with using the tool. In short, the support structures need to be in place – something that will be more important as smartphone membership expands into sections of the population who are perhaps less comfortable with the technology.

3. Can people multi-task? Yes – we had 260 pre-wedding survey responses, 244 afterwards, 783 wedding moments, 218 photos and 505 GPS readings!

4. Did we get closer to a 360 degree view? Here we would give an unequivocal “yes”. The real-time information genuinely took us closer to what was happening. Coupled with the opinion poll and the social listening, it felt like we were playing different research techniques “in the right position”, each tailored to the “task in hand”. This is not to say that there weren’t gaps in the process. In an ideal world we would certainly have wanted to have built in some work with our “App” respondents during the week after the wedding, and to have done a follow-up, nationally representative opinion poll.

In conclusion, we hope that this experience helps researchers to build on lessons we have learned already from other case studies, for example in the world of mobile ethnography and product usage. From our perspective, we have come out of the study with a stronger sense of the place of mobile in research today, and in particular how it can augment traditional techniques and capture previously elusive insights.

THE AUTHORS Simon Atkinson is Assistant Chief Executive, Ipsos MORI, United Kingdom.

Sean Conry is VP, Business Development, Mobile Solutions, Techneos, a Confirmit Company, Canada.