DEGREE PROJECT IN MEDIA TECHNOLOGY SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2018

User acquisition growth hacks for startups

LYDIA NAARDEN

KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE Sammanfattning Den här studien diskuterar metoder som startups kan använda sig av vid förvärvande av nya användare för sina produkter eller tjänster. Studien utgörs av en fallstudie baserat på en startups perspektiv och jämför denna med andra verksamheter. Då många startups är under framväxt söker de kostnadseffektiva metoder för att få sin verksamhet att växa. Företag som Airbnb och använde ”growth-hacking” under sina inledande faser då de saknade resurser. Många bloggar har skrivit om fenomenet, men det saknas ännu vetenskaplig forskning kring vad som är nödvändigt att utföra och ta hänsyn till. Den här studien ämnar bidra med insikter som saknas rörande förvärvande av användare.

Under sex veckor testades flera olika strategier på Locallife. Dessa strategier har sedan tidigare varit beprövade av andra företag. Insamlad data blev analyserad och använd för optimering. Growth-hacking-strategier blev utförda på Facebook för att identifiera vilka metoder som åstadkom tillväxt av användare. Resultaten visade att kalendrar på sociala medier knappt ledde till några konverteringar: efter tre veckor av utförande och observation var antalet nya användare inte tillfredsställande. Däremot var lokalt anpassat marknadsföringsinnehåll effektivt och gav tillfredsställande resultat. Innehåll som användare kan relatera till är mer sannolikt att attrahera användarens uppmärksamhet än generell information om produkten eller tjänsten. Dessutom är den inte enbart kampanjens ton som påverkar åhörare: sättet att uttrycka sig på för att uppmana personer till handling påverkar individers beslut om att ansluta sig till produkten eller tjänsten.

Dessa resultat är betydelsefulla då de erbjuder nybörjare och företag ett perspektiv på vilka frågor och problem som måste tas hänsyn till när de avser att utöka antalet användare. Dessutom, trots att många metoder är diskuterade i bloggar är det ytterst få som förklarar vilka justeringar som genomfördes för att nå ett specifikt resultat. Detta har fått det att framstå som att höga resultat har åstadkommits omedelbart. Den här studien ämnar utforska konceptet growth-hacking och framhäva ansträngningarna som krävs för att en startup ska nå hållbar tillväxt.

2

Abstract This study discusses several methods that startups can use to acquire new users for their product or service, from the perspective of a startup that forms the case study of this research and in comparison to other companies. As many startups are currently emerging, they seek cost-efficient ways for their company to grow. Companies such as Airbnb and Facebook used growth hacking when they started and lacked sufficient resources. Many have been written about this phenomenon, but there is a lack of scientific research examining what needs to be done or considered. This research aims to contribute to filling that gap in relation to user acquisition.

For six weeks, several strategies adopted by companies were tested at Locallife to determine which ones worked. The collected data were analysed and used for optimisations. Growth-hacking practices were performed on Facebook to identify which hacks lead to growth. The results show that the social media content calendar barely led to any conversions: after three weeks of execution and observation, the number of people that onboarded was unsatisfactory. However, local targeting content yielded satisfying results. Content with which the user can associate is more likely to attract the user's attention than general information about the product or service. Further, not only the tone of the campaign makes a difference for the audience; also the call to action influences people’s decision to onboard and the performance of the product or service.

These findings are significant because they offer beginners and companies a perspective on which issues to consider when they decide to growth hack. Also, while many methods are discussed in blogs, only a few explain which adjustments were made before reaching a particular result, making it seem as if everything happened in one go. This research aims to explore the concept and show the efforts that need to be made for a startup to reach sustainable growth.

3

Table of contents

1. Introduction

1.1 Background………………………………………….………………………….………….…………………….6 1.2 Research question……………………………….……………………….…………….…………………….7 1.3 Purpose and objective……………………………….…………………………………………………………………..8 1.4 Delimitations…………………………………….…………………………………………….………………..8 1.5 About Locallife………………………………….…………………………………………….………………..9

2. Theory

2.1 Growth hacking: what is it?...... 10 2.2 The growth process………………………………………………………………………………………..11 2.3 User acquisition………………………………………………………………………………………………12 2.4 Data-driven digital marketing………………………………………………………………..………13 2.5 AIDA analytics…………………………………………………………………………………………...……14 2.6 Scheduled social media strategy……………………………………………………………...……16

3. Methodology

3.1 Strategy…………………………………………………………………………………………………………..17 3.2 Data collection………………………………………….…………………………………………….…....18 3.3 Data-collection framework…………………………………………………………………….………18 3.4 Process…………………………………………………………………………………………………….……..19 3.5 Research errors……………………………………………………………………………………….……..21

4. Results and analysis

4.1 Motivational Monday…………………………………………………………………………………..…23 4.2 Educational Wednesday………………………………………………………………………….……..25 4.3 Inspirational Friday……………………………………………………………………………….…….….27 4.4 Local campaigning……………………………………………………………………………….….………30 4.5 Retargeting locals……………………………………………………………………………………………35 4.6 Experiences of other companies……...….…………………………………………………….….38

4

5. Discussion and conclusions

5.1 Discussion…………………………….…………………………………………………………………….……40 5.2 Conclusion…………………………………………………………………………………………………………………….42 5.3 Further research………………………………………………………………………………………..……43 5.4 Limitations………………………………………………………………………………………………….…..43

6. References 45

5

1. Introduction This chapter defines the situation and the problem to provide an understanding of the relevance of the topic. First, the study concept is explained and the central research question is presented. This is followed by an introduction of the sub-question, together with a brief description of the company in which the research was carried out and a discussion of the study’s purpose.

1.1 Background When investors have been found and the company has been built from the ground up. Often the expectation is that not much of an effort needs to be made; the product will sell itself. However, after a couple of years of no existential growth, feelings of uncertainty rise up. A lot of money has been invested in creating the product, so no money is available to drive the growth. Many startups deal with these circumstances and find themselves in an unexpected situation. Now what? Give up? Or try to secure more funding to start marketing campaigns?

According to Sean Ellis (and many other companies, e.g., Airbnb, PayPal and ), growth hacking is a method worth trying when one has a small budget but big dreams. The term ‘growth hacker’ was first introduced by Ellis in 2010, who defines it as a person who solely focuses on how he/she can increase the number of customers for his/her product. Since then, different hacks have been used by several companies to achieve sustainable growth. The foundation of a growth hack builds mainly on data and creativity, pursued by making a small adjustment and experimenting with it to produce a positive result. The concept is a combination of the product or service, data and marketing, but is mostly driven by the product (Ellis, 2017).

Over the past years, there has been an upsurge of enthusiasm to start one’s own business, especially in the field of technology, although half of the initiatives fail within the first five years (Giardino et al., 2014). This is due to the fact that certain important aspects are overlooked in the business process. First, a strong and a well-performing technical team does not ensure sustainable growth. Second, the mindset that the CEO only needs to direct the company, the engineer is tasked only with coding and the chief marketing

6

officer is only responsible for marketing is the wrong attitude. All parties should be involved in the processes, as the responsibilities are organic in a startup (Giardino et al., 2014). When a company decides to try growth hacking, the duties are not as segmented as in an already established company. Instead, all parties work together to achieve sustainable growth (Ellis, 2017).

Creative marketing

Creative Marketing Experiments & data Automation & engineering

Figure 1.1: Growth hacking implementation (reproduced from Bernd Skiera)

1.2 Research question Several books have been written about the concept of growth hacking in general. They are often written by professionals who have applied different hacks to grow companies and who, based on this experience, outline a situation and briefly explain the result. However, there is a lack of research on how to use growth hacking to specifically acquire more users and describing the adjustments that were made to accomplish specific results. As the number of companies that start and fail grows due to the lack of capital to promote their product, there is a need for more research in this field, identifying and discussing the process more thoroughly.

7

Research question v Which user-acquisition marketing strategies contribute to the growth of startups?

Sub-question v Which of the methods used by other startups lead to an increase in the number of users of Locallife?

1.3 Purpose and objectives This research is relevant because although growth hacking is still a new phenomenon within startups, little research has been done on this topic. The purpose of this degree project is to gain an understanding of the concept and how it works in practice. Further, this paper adds material to the (few) existing articles as it provides a bridge between growth hacking and user acquisition: a field in which there is a lack of literature. The goal of this study is to identify the effect that certain decisions or adjustments have on the growth process by discussing the hypotheses and the actual results of each strategy.

1.4 Delimitations This research focuses on the methods that can be used to increase the number of users and does not discuss user retention. The project aims to focus on the process of user acquisition, rather than on the growth hacking funnel as a whole. This decision was made because various studies recommend focusing on a single goal in starting to practice growth hacking. Further, at the time of research, the company was expected to benefit most from expanding its user base. Additionally, the discussed methods were executed on the social network Facebook. This decision is related to the fact that the experiments were conducted within a time span of six weeks; to ensure the reliability and value of the data, the choice was made to mainly focus on a single platform.

8

1.5 About Locallife Locallife is a social network for neighbours. The platform offers an infrastructure for neighbours to get in touch with each other and profit from each other’s social capital. The concept is in line with the economic system of the sharing economy, whereby the sharing of goods and services is encouraged in order to accomplish more sustainable living. Currently, several areas in Stockholm have been activated and the goal is to expand as the company continues growing. The company also started offering services to housing and real estate companies.

Figure 1.2: Newsfeed of Locallife in Norra Djurgårdsstaden

Figure 1.3: Events in Norra Djurgårdsstaden

9

The idea began as a research project by KTH innovation and led to a venture opportunity. The company has existed for three years.

2. Theory This chapter discusses previous research in the field of growth hacking and data analysis to develop an understanding of the topic.

2.1 Growth hacking: what is it? Growth hacking is a form of social media marketing, though it is not the same as marketing. The difference between these two concepts is that marketing puts the promotion and selling of an already created product in perspective, while growth hacking tactics entail testing the different elements of a product or service on consumers to acquire data that can be analysed and subsequently improving the product so that sustainable growth can be achieved (Skiera, 2016). To achieve growth, the service-centred view of marketing applies, meaning that people are active in the product’s value-creation process (Vargo & Lusch, 2004).

By using growth hacking tactics to grow, a company utilises its resources more efficiently by reallocating them to the areas where they are needed for higher growth conversions. For startups, resources are already scarce, so being able to divide them into different focus areas is a matter of importance (Reinartz, Thomas and Kumar, 2005). By adopting this concept in the company’s early stages, the ‘field of dreams fallacy’ is addressed, which refers to the mindset that as long as the product provided is excellent, people will use it (Fong and Riddersen, 2016). While this attitude worked in a particular time in history, people currently have more options, there is more competition and customers may not know they need a certain product. This entails that the necessity needs to be created by the creator, which is why knowing how or when to present the good to the consumer is crucial for the growth process to be successful (Blake, Nosko and Tadelis, 2015).

Growth hacking can be done in different ways, as long as the process entails a combination of data and adjustments in both the content and the product or service. Some examples of growth hacking tactics are A/B testing, paid acquisition, and Search Engine Optimization (SEO) (Ellis, 2017).

10

It is easy for startups to get caught up in trying to achieve growth in every part of the business. However, according to the authors of Hacking Growth, it is better to focus on a single metric and grow it very well. If the company’s engagement is very low but it has a high number of users, it may be better to focus on engagement by experimenting with different methods. This prevents the company from locking up substantial amounts of working capital or ending up in the ‘data jungle’ (Skiera, 2016).

2.2 The growth process As mentioned above, before the process starts, it is important to define the area on which the company will focus its resources to avoid doing too much, which ultimately leads to not doing anything. The idea behind growth hacking is to follow a process that continues until the right strategy for achieving growth is found. It starts with proposing ideas for strategies that may lead to achieving the expected growth. Often, these ideas are strategized by the marketer, after which the rest of the team has the opportunity to add input. However, Ellis, the creator of the term ‘growth hacker’, recommends involving the team in the process as early as possible. After the best ideas are chosen, the testing period can start. Based on the results, the trick is to identify the bottleneck (‘analyse’) and subsequently make small adjustments (‘optimize’) to reach the best results until the set goal is reached. In this process, the learning experience is documented to avoid repeating the same mistakes. It is possible to run several experiments at once so that as much data as possible are collected to identify trends. However, it is recommended to establish a dashboard, which only tracks the needed data (Ellis, 2017).

11

1.Ideas

5.Optimize 2.Prioritize

4.Analyse 3.Test

Figure 2: The continuous growth hacking process (adapted from Sean Ellis)

2.3 User acquisition The time when users used a product because it was marketed (pushed) to consumers is a thing of the past. Consumers want to know what the value of the product is for them, and the content needs to have relevance for them in real-time and be viewed on different channels for it to influence consumers (inbound marketing). To be able to do this, it is essential to understand customers before they have their first encounter with the product. On which social media platform are they active? At which time? Why? And what are they interested in seeing or reading? This helps to obtain a successful strategy to increase brand engagement (Lemon, 2016).

How much a company should spend on acquiring new users differs. A fully established company is likely to have more money to spend and be able to run ads on TV, while for startups with a limited amount of financial resources, it is essential to make user acquisition efforts as effective as possible. In starting to develop a strategy, to prevent large sums from being spent on ads, it is recommended to start with investing small amounts in several tests and increase the budget of the strategies that yield results. Because of startups’ lack of resources, their growth process is often far slower in comparison to that of larger companies, though as time passes the benefits and the growth drivers increase (Ellis, 2017)

12

Facebook is often used by startups to spread the word about their products, and it has the advantage that it offers many different ways to create content and target a specific group. Additionally, it provides the possibility to try A/B testing and (can) show results after a couple of days. However, the marketing channel should be chosen depending on the target group (Wilken, 2014). It often happens that the initial target group does not comprise the people who are interested in the product. This can be determined by examining the data when the experimentation process starts. However, this does not mean the initial target group cannot be reached, but identifying why the people who were first targeted were not interested in the product can offer insights on what to do differently next time. In the process of acquiring users to the platform, it is essential to ask the question ‘why’ at every step of the cycle. It is also important to bear in mind that the people who are initially considered to be interested in the product are not necessarily those who will be reached. However, through the data acquired during the process, the target group can be targeted (Jansen, Menichelli and Næs, 2015).

2.4 Data-driven digital marketing Over the past few years, data have been of significant relevance for businesses. Especially technical companies are highly data-driven as they use data to determine which action needs to be taken next or what could happen in the future: a process called ‘forecasting’ (Provost and Fawcett, 2013). Because of the growing importance of data, there are different tools to gather data (Amplitude, Mixpanel, Analytics, etc.). Often, a combination of different tools is used to create a valuable dashboard. In growth hacking, every decision that is made is based on data. Data therefore becomes a very important factor for the company and hence analytical skills and a dashboard are needed for the decision-making process. However, as data themselves do not mean anything, a testable hypothesis needs to be formed based on the data. The concept of data-driven marketing uses already established information on consumers for optimisation and targeting purposes, making it possible to target customer more efficiently. Hence, already established companies that have already reached a product-market fit mostly use this form of marketing (Järvinen and Karjaluoto, 2015). Data-driven digital marketing is discussed in this research in view of indicating the differences between this concept and growth hacking, one of the most important being that growth hacking is based on experiments that run for a short term and with a small

13

budget. Further, while pursuing growth hacking, the company is often still in the phase of improving its service or product. In contrast, with data-driven digital marketing, a product or service that has already been tested and ‘approved’ by consumers is marketed to the target group. Hence, when a company uses growth hacking, the consumers are operand resources, meaning they are the ones who produce the effect, rather than the ones who are affected (Vargo and Lusch, 2004). It is important to remember this, as considering this concept as ‘only’ marketing is wrong and does not do it justice. Data-driven digital marketing can be used while growth hacking, but only marketing the product to the consumer is not growth hacking unless adjustments are made in the content and the product or service that are offered. This makes the process dependent on technical, marketing and data-driven expertise (Tiago and Veríssimo, 2014).

2.5 AIDA analytics In using social media as a growth strategy, it is essential to have a tracking system to track the needed funnels. Attention, Interest, Desire and Action (AIDA) analytics is mostly used and has been recommended by many scholars as an excellent marketing-analytics framework to optimize the user experience. It describes the steps through which a potential customer goes before deciding to commit to a product or service. The model is used in marketing and advertising to explain the steps or stages from the very first moment a consumer is aware of a product to the moment in which action is taken (Rawal, 2013).

The framework goes through the different stages of the customer journey using the following funnels: v Attention (awareness): This explains how well the strategy or the used marketing channel is performing (e.g., the number of people who clicked or engaged with the content). An element that influences these results is the location where the ad is placed. On most social media platforms, the placement can be selected. However, the recommendation is to let the social media platform automatically choose the ad’s placement if the product’s target group is broad or if there is no information on the preferences of the targeted people. According to Fan and Gordon, this increases the reach and likelihood of capturing people’s attention (Fan and Gordon, 2014).

14

v Interest: This refers to whether people like the content or website that is shown (Heimbach and Hinz, 2016). For instance, if the consumer visits the platform (attention), but immediately drops off, this is often signified as ‘bounce rate’ in analytic tools. Usually, this means there is something on the website that has become a bottleneck for the conversion (Anderl et al., 2015). v Desire: In this step, it is essential to tell the consumer about the benefits (for them) of the product as this is the last step they go through before they make a decision. This can be done both on the platform and in the ad, meaning the framework does not always follow all the steps. Depending on the landing page, a decision can be made about which content the ad or post will contain. Regardless of the decision that is taken, it is relevant to track how the consumer deals with it to establish a pattern of what works and what does not (Morin, 2011). v Action: In this stage, the consumer becomes a customer/member, also called ‘closing the sale’. This result can be achieved by showing small triggers on the website or having an ad with a direct call to action (CTA): ‘sign up’ or ‘register’.

Depending on the metric, all four steps can be shown in an ad/post or the marketing channel can be used as a trigger (Rawal, 2013).

The abundance of data offers companies the possibility to tap into the information of a large crowd, which helps them to make more informed decisions. However, the wealth of data also makes it difficult to determine what to pay attention to. A goal needs to be identified first, after which the focus points need to be established in advance, based on which hypotheses need be formed that can be tested and rephrased until the desired result is achieved (Liben-Nowell and Kleinberg, 2007).

15

2.6 Scheduled social media strategy According to Geho and Dangelo, over 70% of entrepreneurs find the use of social media critical to their business, as it leads to higher exposure for their business. Scheduling content or ads at specific times and days increases the consumer’s engagement as this sets an expected standard on which the consumer is relying. Several authors recommend entrepreneurs to start with a scheduled marketing strategy, as this is a very successful way of obtaining followers and leading them to become hooked on a product. However, it is vital that the produced content not only promotes the product but also sympathises with the consumer (i.e., fun, educational and motivational content). Yet, it is essential to create content that is relevant for the target groups and has a purpose. Inserting a CTA in an ad is a highly recommended way of increasing the conversion rate. However, it is crucial to not only create content with the CTA ‘register now’ but also to alternate it with CTAs such as ‘learn more’. This shows the audience that they are valued and appreciated by the brand, regardless of whether they have made up their mind about taking action or are still in the ‘desire’ stage (Geho and Dangelo, 2012).

16

3. Methodology In this chapter, the methodology of this thesis is described and explained in detail. Additionally, errors that may have occurred during the testing period are discussed in correlation with their effect on the results.

The growth process of Locallife was followed for four months and a half. The task was to implement different strategies that may lead to growth in the number of members on the platform. For six weeks, several tests were executed with the goal of increasing the brand awareness of Locallife. For each test, a hypothesis was formulated based on previous research or results yielded during the experimentation period. Google Analytics and Facebook Pixel were used to collect the data that are presented in the results.

The ads and posts were designed by the author and discussed with others, with whose approval the content was published. Further, as the platform targets Swedes, the material was in Swedish. As I am not a Swedish speaker, a team of native Swedish speakers was made available, who translated the texts from English to Swedish.

3.1 Strategy Facebook was the primary channel used to reach the audience; however, as it also has the feature to post on , this platform was also used. The experiment started with a marketing strategy based on a schedule: the social media calendar. Three days a week, on Mondays, Wednesdays and Fridays, different content was posted on Facebook, each of which had the goal of creating brand awareness and increasing the conversion rate. Monday was for ‘motivational’, Wednesday for ‘educational’ and Friday for ‘inspirational’ content. The same schedule was used for the ‘This Is Nuts’ campaign (further described in section 4.6). Because of the unsuccessful results of the social media calendar strategy after three weeks, the decision was made to make changes on the platform (Felix, Rauschnabel and Hinsch, 2017). Further, the experiences of other companies using growth hacking strategies are discussed in the results section. Their experiences and the results of their experiments are also addressed in the discussion section to make a comparison.

17

3.2 Data collection This study is based on quantitative research as it offers the possibility to measure data using statistics and is more objective, as the connection between variables is studied based on an analysis. However, this method does not consider the context of the experiment, which entails meanings are based on the researcher’s interpretation. This study aims to interpret the material based on numbers, which is why the quantitative method was chosen. Further, the meaning of the audience’s actions is not particularly relevant for this study and hence is not included.

For six weeks, several social media strategies were executed on Facebook with the purpose of acquiring new users to the Locallife platform. During the first three weeks, the ‘scheduled social media’ approach was used as an acquisition strategy. According to Hoover, it is important to set an expectation, and the best way to do this is by adopting a routine. The digital branding agency Digital Fans (This Is Nuts campaign) adopted a framework in which they posted three times a week, with every day having a different goal. The same structure was used for Locallife to increase engagement and brand awareness. After three weeks, the decision was made to change the strategy due to circumstances that are discussed in the results section. Every approach was based on a hypothesis that was tested, and the results either validated or invalidated the assumption.

3.3 Data-collection framework The marketing strategies were tracked using AIDA analytics. This system was chosen because it is widely used by growth hackers and because various studies recommended it. In the results section, the four components of AIDA are shown as follows: v Attention: Reach, i.e. the number of people who saw the ad or post v Interest: Engagement, i.e. the number of people who liked, shared or commented on the post v Desire: Clicks, i.e. the number of people who clicked on the ad or post to visit the landing page v Action: Conversions, i.e. the number of people who created an account (not included in every post as conversion was not the goal)

18

This model offers the opportunity to stay focused on the set metrics and requires little time to understand. According to Bijmolt, the problem with many analytical frameworks is that they are complicated, which makes the statistics unknown to the managers and the rest of the company. He proposes to develop standardised models in which the relevant variables are fixed and transparent. However, the disadvantage of this model is that when the metrics change, the model needs to be reconsidered as well and probably also the analytics dashboard, as the right metrics may not have been tracked (Bijmolt et al., 2010). The decision to use AIDA analytics was in consideration of the time span of the experiments and resources. As the experiments were executed over six weeks with the goal of acquiring new members to the social network, the disadvantage of the model did not influence the results significantly. Notwithstanding, not all of the metrics were tracked by the strategies.

The collected data are based on a hypothesis and therefore subject to bias, making the focus rather narrow. Additionally, several of the decisions made were based on an individual point of view and hence subject to personal interpretation. However, to ensure the validity, experiments were executed based on the hypothesis, and the assumptions were often discussed with the rest of the team and adjusted according to the obtained feedback. Regardless of the measures taken to decrease the level of subjectivity, the results remain subject to interpretation.

3.4 Process The continuous growth hacking process took place during the six-week experimentation period, meaning the most effective and operable ideas from a series were prioritised and tested during this period. A number of findings from the data were selected to generate new hypotheses and measures over a certain period, until the adjustments no longer led to significant result. When this occurred, the product was modified and tested again, and the results were evaluated and modified as needed. As previously mentioned, the new insights are mainly based on the author’s observations and interpretations. However, the steps taken are based on the supporting literature as well as a number of assumptions.

19

As the company at the time of this research was in the process of activating new areas and opportunities in the market, the decision was made to try them out on social media to obtain responses from the audience. These hacks were tested online for a short time span and proceeded offline.

Prior to this process, an empathy map was generated to obtain a view of the audience’s needs and values. The experiments took this into consideration as much as possible, however not all requirements were met due to the lack of recourses.

Figure 3: Empathy map Locallife

20

3.5 Research errors Every research has faults that influence the data, the data collection and the measurement method. These are identified in this section in view of bearing them in mind in interpreting the results.

Biases: Biases are errors that intentionally or unintentionally occur due to the researcher’s interpretation. As the researcher had the intention to show which strategies are successful, her biases may have inadvertently influenced the study. Hence, the results may present her preferences for a positive outcome.

Measurement errors: These errors, which are present in all studies, indicate the difference between the actual meaning of the data and the researcher’s interpretation of them. One type of error that may have occurred during this research is random error. For every strategy, the data were analysed after seven days, with the goal of making the measurement at 10:00 AM. However, this may have occasionally deviated by a couple of minutes, leading to inconsistencies in the quantities.

Data analysis errors: This error occurs due to the researcher’s interpretation of the data. The data were collected with two different systems: Google Analytics and Facebook Business manager. Any mistakes made in the process affect the drawn conclusions (Vogt, 2011).

21

4. Results and analysis This chapter shows the results that were collected during the experimentation period and analyses them to explain the optimisations. Further, an analysis of the data is provided to illustrate the effects of the data adjustments and indicate the stage at which a positive result was achieved. Every experiment has hypotheses that are tested; these are also presented in this section.

Three experiments were executed, each of which had a hypothesis. During the tests, several adjustments were made in the campaign or product to reach better results.

The first test is based on the social media calendar. For three weeks, different content was pushed on Mondays, Wednesdays and Fridays, each of which had a specific theme. The aim of this test is to accomplish consistency in view of creating an expectation among consumers. This test is based on the calendar used for the ‘This Is Nuts’ campaign. The campaigns were published at 12:00 PM because that is when most people take a lunch break, at which time the likelihood that they look at their phone is high. The two other experiments are based on the experiences of other companies (see section 4.6). After the ensuing discussion of the tests, a list of the findings can be found. These are considered in the subsequent experiments as they form part of the learning process and increase the likelihood of achieving growth. The second experiment, the local landing page, has two hypotheses because adjustments needed to be made after the first test to establish the ‘hack’. The third test targets the new areas and a more active target group than those targeted in experiments 1 and 2.

Hypothesis 1: Adopting a scheduled social media strategy leads to an increase in brand engagement. Hypothesis 2: A local landing page leads to a higher conversion rate. Hypothesis 3: A local picture on a local campaign leads to higher conversion rate Hypothesis 4: Local information specifically targeting the area of residence leads to higher engagement.

Goal • To attract locals to the Locallife social network

22

Key performance index (metrics) • How many people were reached? • How many people engaged with the content/product? • How many people created an account?

4.1 Motivational Monday

Hypothesis: Adopting a scheduled social media strategy leads to an increase in brand engagement.

Campaign 1 Campaign 2 Campaign 3

Translation of campaigns Campaign 1: The love of our neighbour in all its fullness simply means being able to say ‘Hi, how are you doing today?’ Campaign 2: I always thought I had good neighbours. It turns out I have the best neighbours in the world (text on the picture). Tag your neighbour to show your appreciation. Campaign 3: The duty to help oneself in the highest sense involves helping one’s neighbour (text on the picture). What do you think? When did you last help a neighbour out?

23

Total reach of campaigns 1-3 Reach

1800 1600 1400 1200 1000 800 600 400 200 0 Campaign 1 Campaign 2 Campaign 3

Figure 4.1: Number of people reached with ‘Motivational Monday’

Total engagement of campaigns 1-3 Engagement

8 7 6 5 4 3 2 1 0 Campaign 1 Campaign 2 campaign 3

Figure 4.2: Number of people who engaged with the campaign (likes, comments, shares and clicks)

As is evident from figure 4.1, the reach of the campaign increased on a weekly basis, meaning the number of people who saw the post or ad and contacted the brand grew. However, the number of people who engaged (figure 4.2) with the post was meagre. This trend is probably due to the content not being of interest to

24

the target group, but may also be connected to the company’s lack of social media presence.

4.2 Educational Wednesday

Campaign 4 Campaign 5 Campaign 6

Translation of campaigns 4. Did you know that Norra Djurgårdsstaden is one of Stockholm's most environmentally developed areas? 5. Did you know that Sundbyberg was detached from Bromma in 1888 as a market town? 6. Did you know that in the 1950s Bagarmossen received international recognition for its consistent separation of walking and road traffic by installing many footpaths and that the houses follow the nature?

25

Total reach of campaigns 4-6 Reach

6000

5000

4000

3000

2000

1000

0 Campaign 4 Campaign 5 Campaign 6

Figure 4.3: Reach of ‘Educational Wednesday’

Total engagement of campaigns 4-6 Engagement

60

50

40

30

20

10

0 Campaign 4 Campaign 5 Campaign 6

Figure 4.4: Number of people who engaged with the campaign (likes, comments, shares and clicks)

As is evident in figure 4.3, the campaigns had a large reach. What is striking in figure 4.4 is the high level of engagement, especially in comparison to campaigns 1- 3. This finding shows the target group appreciated being provided with information about their neighbourhood, which complies with the needs identified in the empathy map (see figure 3). With this insight, the decision was made to proceed

26

with facts and information about different neighbourhoods, targeting the people who live there.

4.3 Inspirational Friday

Campaign 7 Campaign 8 Campaign 9

Translation of campaigns 7. Ask your neighbour to walk your dogs together and have a discussion about what you could do to make your neighbourhood better or about organising a community event. Share your ideas on Locallife! Not a member yet? Click on the sign-up button. 8. What do you think/feel should change in your neighbourhood? Share your opinion in the comment!’ 9. Invite neighbour for lunch this weekend and make plans on how you can contribute to a change in your neighbourhood. Is he/she not a member of Locallife yet? Invite them to join the social network to see who their neighbours are!

27

Total reach of campaigns 7-9 Reach

1000 900 800 700 600 500 400 300 200 100 0 Campaign 7 Campaign 8 Campaign 9 Figure 4.5: Number of people who saw the campaign

Total engagement of campaigns 7-9 Engagement

180 160 140 120 100 80 60 40 20 0 Camaign 7 Campaign 8 Campaign 9

Figure 4.6: Number of people who interacted with the campaign (likes, comments, shares and clicks)

Figure 4.5 shows the reach of campaigns 7-9. What is interesting in figure 4.6 is the high rate of campaign 9. This can be related to the use of pictures, as content about food often performs well on Facebook. However, this can also be a pattern, as the previous campaigns (5 and 1), which also had a ‘learn more’ button, also performed well.

28

This Is Nuts used a similar content calendar to improve the company’s social media presence (see figure 4.13). It worked out very well for them, and other researchers recommended using this strategy. As is evident, the reach of the campaign is large and the engagement is satisfactory. Surprisingly, the conversion was not satisfactory. During the three weeks, only 10 people became members of the platform through social media. The analysis of the bounce rate and session duration led to the realisation that the session duration was in order: new visitors spent an average of 2.26 min on the page. However, the bounce rate of the landing page was 78%, meaning most people only viewed that page and did not look further, exiting the site without interacting. These data show the strategy was not successful in converting people, nor in creating the high level of engagement that was expected at the beginning. However, it did yield data that were used in the second experiment.

Findings from experiment 1 1. The ‘learn more’ button, a CTA, leads people to engage with the campaign 2. Information about a neighbourhood targeting the people who live in that area also leads to an increase in engagement 3. The campaigns reached a consistently large audience between 4:00-9:00 PM

Audience activity for campaigns 1-9

Figure 4.7: Total audience reach per hour for campaigns 1-9

29

What is interesting in figure 4.7 is the daily audience activity. As is evident, the audience mostly engaged with the campaigns from 4:00-9:00 PM. This can be explained by the fact that most people spend time on leisure and social media after working hours, which increases the likelihood that they see the content. This information was considered in the subsequent experiments.

Due to the unsuccessfulness of the campaigns in converting and given the high bounce rate, the decision was made to create a new landing page customised for a specific area, with the assumption this would be more successful.

4.4 Local campaigning

Hypothesis: A local landing page leads to higher conversion. This is based on the result of the high bounce rate and the high level of engagement of the local campaigns. *

Figure 4.8: Locallife local landing page Norra Djurgårdsstaden

* The hypothesis was tested on the area Norra Djurgårdsstaden 30

Figure 4.9: Regular landing page of Locallife

Campaign 10

Figure 4.8: Audience campaign 10

31

Translation of campaign 10 400 neighbours in Norra Djurgårdsstaden have just joined the local social network Locallife—see who is involved and get to know your neighbourhood and your neighbours a bit better.

Total results of campaign 10

Results 1269 1400 1200 1000 800 600 400 200 47 9 0 Reach Clicks Conversions

Figure 4.9: Results of campaign 10

Although the numbers in figure 4.8 may still seem low, considering the ad ran for one week, the result is satisfactory. The fact that 9 people out of 47 converted is a positive sign, meaning the change of the landing page made a difference. Further, the bounce rate of 78% decreased to 75%, which may not sound like an accomplishment either, but it surely is. Growth hacking is about the small wins, which are seen as accomplishments, as ‘growth’ takes time. Nine people became members of the platform through this campaign, which is not sufficiently high to achieve sustained growth, but is a step in the right direction. However, the number of clicks was still low in comparison to the reach, so the decision was made to change the picture of the campaign (see campaign 11) to something more local, while keeping the text in view of measuring the effect of the picture change.

32

Hypothesis: A local picture on a local campaign leads to higher conversion rate

Campaign 11: Increasing conversions in Norra Djurgårdsstaden

The picture made a difference in the conversions: at the end of the week, 12 people had registered on the platform, and the bounce rate remained unchanged. This result is good, but still not satisfactory, so the decision was made to adjust the target audience by narrowing down the number of interests.

33

Figure 4.10: New audience of campaign 11

Figure 4.10: New audience campaign 11

The audience adjustment led to a weekly increase in the number of new registrations, varying between 15-20 new registrations per week. This can be viewed as a steady growth for the company. The bounce rate of people who arrived at the landing page through the social media campaign decreased to 71% and the average session time increased from 2.26 to 2.45. This hack was so successful that the company decided to keep running the campaign to sustain its growth.

As mentioned in the previous section (4.3), several points stood out during the preceding campaigns: 1. Using CTAs yields more results 2. Local information leads to higher engagement 3. Campaigns reach a peak between 4:00-9:00 PM

With this information, more targeted campaigns were created for residents living in a new to-be-activated area, as well as for Locallife leaders and real estate companies.

34

4.5 Retargeting locals

Hypothesis: Local information specifically targeting the area's residents leads to higher engagement

Campaign 12 Campaign 13 Campaign 14

Translation of campaigns 12-14 12. Are you on the board of a housing company in Norra Djurgårdsstaden? Are you looking for a way to make your tenants more comfortable while at the same time reducing your workload? Then we have the solution for you! Locallife provides a service for housing companies to make the communication with your tenants easier, store documents and get in contact with other housing associations. Your tenants will have the opportunity to get to know each other, establish a community-sharing economy and many more features! We will help you and your members to get started, and we are available to answer your questions. You get all of this for free! Locallife, the digital tool that contributes to local development, as well as ecological and social sustainability. Strengthen the community. Improve administration. Fully free of charge. Click on the ‘sign-up’ button and get started now!

35

13. We are looking for Locallife leaders in Lidingö! Do you have a passion for making your neighbourhood more sustainable? Then join the movement and become the Locallife community influencer! 14. Did you know that Aspudden was previously divided into Large and Small Aspudden? During the 19th century, there were flowers and fruit cultivars at Small Aspudden, and at Large Aspudden, tobacco was grown.

Total reach of campaigns 12-14 Reach

9000 8000 7000 6000 5000 4000 3000 2000 1000 0 Campaign 12 Campaign 13 Campaign 14 Figure 4.11: Amount of people reached with campaigns 12-14

Total engagement of campaigns 12-14 Engagement

60

50

40

30

20

10

0 Campaign 12 Campaign 13 Campaign 14 Figure 4.12: Amount of people who engaged with campaigns 12-14

36

What is evident in figures 4.11 and 4.12 is that a large reach does not necessarily qualify for high engagement. The results reveal that although the scope of campaign 14 was low, the engagement was higher than in the other campaigns. Also evident is the low level of engagement of campaigns 12 and 13. The assumption is that the campaigns targeted different kinds of people: people who wanted to do more than be a member of the social network and housing companies. Hence, a different social network (e.g., Twitter) would be more appropriate to reach these people.

The company started using Twitter, and several Locallife leaders and housing companies showed interest in the service. Additionally, a significant amount of manual work was done to reach out to these people in order to get them on board.

Experiment Hypothesis Results

The social media calendar Adopting a scheduled social media strategy leads Large reach but low to an increase in brand engagement. engagement: 10 registrations after three weeks

The local landing page 1. Having a local landing page leads to an increase 9 registrations in one week in conversions.

2. Posting a local picture on a local campaign leads 15-20 registrations in one week to a higher conversion.

Retargeting locals Local information specifically targeting the area’s Small reach but high engagement residents leads to higher engagement.

Figure 4.13: Overview of results

37

4.6 Experiences of other companies

This Is Nuts The company This Is Nuts applied a marketing strategy centred on motivating, educating and inspiring its audience to increase brand awareness among a new audience. This was implemented three days a week, with every day having a different topic-engaging content for publication. The brand chose to use this strategy because it wanted to let people know about its existence by conveying a clear brand strategy. As a relatively new company, it tried to communicate what it stands for: nuts made with only natural ingredients without sacrificing taste. This strategy yielded significant results, as its number of followers at the end of the campaign had increased by 35% and its brand engagement was 14.8 times higher than the industry average (Digitalfans.se, 2018).

This Is Nuts scheduled marketing strategy

Figure 4.13: Social media campaign This Is Nuts (reproduced from Digital Fans)

Calroo family The company offers an application on which families with a busy schedule can organise their tasks to attain an overview. When the company started using Facebook, it often shared information about the service but this information was not necessarily relevant for the user. For instance, it shared general information about how parents globally dealt with their busy schedules. Because this information did not provide value for people on an individual level, they reached over 3,000 people and had between 1-5 engagements. When the company started

38

focusing on content with which the audience could associate on a more national level, its reach was around 2,000. However, its post-engagement was about 200. The company did not spend more money on the campaign, but rather invested more effort in relating to its audience (Foundr, 2018).

NomadApp The NomadApp is an application for backpackers on which they can find destinations that are available for a reasonable price. Its social media strategy is based on the following formula: a question or a statement to make the audience curious + stating the benefit for the audience + CTA = success. With this strategy, the company managed to acquire thousands of likes within six months. Before they arrived at the point of success, they tried different strategies and made small adjustments until they yielded this result. This example also shows the relevance of ‘knowing your audience’ as that is the only way relatable content can be created (Foundr, 2018).

39

5. Discussion and conclusion In this chapter, the main findings are discussed in relation to the theory. The questions posed in the Introduction are also answered, a recommendation is offered for further research and the limitations of this research are discussed.

5.1 Discussion Over a period of six weeks, several social media frameworks were implemented to drive the growth of Locallife. These were also based on structures used by other companies. The results of the experiments can be found in chapter 4. The findings of the tests suggest that a framework’s success in one situation did not assure its success in another. The results in sections 4.1-4.3 indicate the scheduled social media strategy used “for the campaign of This Is Nuts” did not yield the same positive effect for Locallife. A likely explanation for this is that the information in the campaign was too general and hence not perceived as unique by the audience. Further, the promoted product also influences the likelihood of the campaign’s success. Regardless of how nicely the content is designed, if the product features do not comply with the needs of the target group, this is likely to impact the engagement/conversions as well, which means it is likely that adjustments in the product influence the performance of the social media content.

Another finding regards the number of adjustments that need be made to reach sustainable growth and how all these (small) changes influence the overall goal. As is evident in section 4.4, the changes in the landing page led to an increase in the number of registrations on the platform, not only on a general level but also regarding the amount of people onboarding on weekly bases. The decision to create a new landing page was based on the data showing that the landing page was not functioning as intended. It is almost certain that the (local) content in the ad and on the landing page was more satisfactory to the audience. However, another probable explanation is that there were only two buttons (the general page had three) on the landing page, making it easier for users to know what is expected of them. Several blogs explain different ‘tricks’ that can be executed to drive user growth and sometimes suggest that imitating them will lead to a similar result. However, this research shows that the product influences the impact. Reproducing an

40

already established framework does not guarantee success. Frameworks need to be adjusted to the product for them to yield any result.

There are some similarities between the experiences of Locallife and those of other companies. As explained in section 4.6, a more narrow approach leads to a higher result, even if the target group is smaller. The Calroo family saw its number of engagements increase by 100% only by providing content with which the consumer could associate, instead of providing information on a more general level. The same trend was identified with Locallife when the content was narrower, hence reaching a smaller group. The most likely reason for this is the tendency of these people to already have an interest in their area, which increases the chance they will engage with the content.

It is not always possible to create content customized for a particular group, especially as the company was expanding its reach at the time of the research. However, when a company is in its startup phase, the recommendation is to start small and test a strategy on a small group, and to learn from the experience and take the success factors to the next group. In the case of Locallife, the company could bring this experience to the following area they want to activate (Fong and Riddersen, 2016).

Another similarity can be identified between NomadApp and Locallife. The approach used by this company is similar to that executed by Locallife in campaign 11. As mentioned above, NomadApp implemented the following strategy: posting a question or statement to make the audience curious + stating the benefit for the audience + CTA = success. These attributes were present in campaign 11, which was the most successful campaign for the company. A probable explanation for this is that, instead of the campaign describing what the platform does, it explains an advantage for the user, which is what consumers are looking for: something to make their lives easier. Nir Eyal, the author of Hooked, also explains that putting the consumer first instead of the product makes a big difference for the audience (Eyal and Hoover, 2014). Again, this is also related to the audience’s ability to make associations. However, Campaign 13 checks all these boxes, though its engagement was low. This slightly contradictory result may be due to the lack of brand awareness in that

41

neighbourhood, as the service is not yet active in the area. It may also be related to the campaign or the platform that was used not appealing to the target group.

It is clear that finding a method that contributes to the growth of a company is a step-by-step process and does not happen overnight. Several changes need to be made for adjustments to yield results.

5.2 Conclusion Which user-acquisition marketing strategies contribute to the growth of startups? For startups, it is a matter of relevance to target a specific group, which means the reach will be small though the results will be better in comparison to strategies that target the general customer. This should not only be done with the campaigns; also the product or service should be compatible with the target group. When the organization is in its beginning stage and its customer base is not particularly large, it is easier to customise, hence the company should offer this service as much as possible. This makes the early adopters appreciate the effort and be more likely to spread the word, which is beneficial for the company’s branding. In a later stage, when customization is no longer possible due to the large amount of members, the choice can be made to change the approach. But in the beginning, it’s important to offer the consumer that extra value. Further, setting an expectation by being consistent can also drive growth. However, this is highly dependent on the product or service the company provides, as well as the content.

Which of these methods lead to an increase in the number of users of Locallife? The success of Locallife’s user-acquisition methods can be attributed to different aspects which may seem to be minor changes, but contributed to the performance of the campaigns. The evolution of the landing page from a general to a local page made a significant difference in the platform’s bounce rate. The decrease of 3% was the first sign of achievement in the process. Because the result still did not meet the expectation, the decision was made to change the picture of the campaign, as a connection was made between the results of the local landing page and the use of a local image in the campaign. As mentioned in the results, this led to 12 new registrations by the end of the campaign week. As the experimentation period spanned six weeks and three of them had already been used for the social

42

media calendar strategy, the decision was made to let campaigns run for a shorter periods so that the data could be analysed and adjustments could be made in time. This decision was made especially in consideration of the importance of changing only one aspect in making adjustments to campaign, for instance age, so that it is apparent which of the elements influence the growth. To achieve more sustainable growth, the decision was made to narrow down the audience, making it easier to reach people who may be willing to onboard and share an interest that complies with the values of the platform. This change is what eventually made the most significant impact, leading to 15-20 new registrations on a weekly basis, with the hours after 4:00 PM being the time in which most people registered. A probable explanation for this is that people then have more time to go through their Facebook feed after working hours and close to dinnertime, making it more likely that the target group sees the ad and converts.

5.3 Future research An area for possible future research is to investigate the reasons why specific changes or strategies have a better effect than others and to research the field on a larger scale by executing not only user-acquisition strategies but also user- retention to conclude on a more extended range. Further, it would be interesting to observe a company for a longer time span to determine for how long the ‘hack’ is effective. Also, due to the short time span, the number of strategies that were executed was small. In future research, more strategies could be executed over a longer time span to formulate more reliable recommendations. Finally, more key performance indexes could also be measured to collect more data.

5.4 Limitations The frameworks used in this research were adjusted to be compatible with the purpose of the studied company, so that the same changes could be made in replicating the experiments. Further, during the time this research was executed, there was a lack of prior research on this topic, which led to the foundation of the problem being based on real-life problems and a small amount of literature. Additionally, due to the short time span of the experiments and data collection, the reported data only show the growth during that timeframe. No conclusions can be made about the methods’ longitudinal effect on the company’s growth. Another limitation that should be considered is that the data is self-reported; there were no

43

pre-existing data to make accurate comparisons. It is difficult to make generalisations based on these results, as they are based on a single case study. Hence, the drawn conclusion can be regarded as a working theory.

44

6. References

Books Ellis, S. (2017). Hacking Growth. The Crown Publishing Group. Eyal, N. and Hoover, R. (2014). Hooked. Fong, R. and Riddersen, C. (2016). Growth hacking. Vogt, W. (2011). SAGE Quantitative Research Methods. London: SAGE Publications.

Articles Anderl, E., Becker, I., Wangenheim, F. and Schumann, J. (2015). Mapping the Customer Journey: Lessons Learned from Graph-Based Online Attribution Modeling. SSRN Electronic Journal.

Bijmolt, T. H. A., Leeflang, P. S. H., Block, F., Eisenbeiss, M., Hardie, B. G. S., Lemmens, A., & Saffert, P. (2010). Analytics for Customer Engagement. Journal of Service Research, 13(3), 341-356.

Blake, T., Nosko, C. and Tadelis, S. (2015). Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment. Econometrica, 83(1), pp.155- 174.

Fan, W. and Gordon, M. (2014). The power of social media analytics. Communications of the ACM, 57(6), pp.74-81.

Felix, R., Rauschnabel, P. and Hinsch, C. (2017). Elements of strategic social media marketing: A holistic framework. Journal of Business Research, 70, pp.118-126.

Giardino, C., Wang, X., & Abrahamsson, P. (2014). Why early-stage software startups fail: a behavioral framework. In International Conference of Software Business (pp. 27-41). Springer, Cham.

Heimbach, I. and Hinz, O. (2016). The impact of content sentiment and emotionality on content virality. International Journal of Research in Marketing, 33(3), pp.695-701.

45

Jansen, J., Menichelli, E. and Næs, T. (2015). Modeling target group heterogeneity in experimental consumer studies. Food Quality and Preference, 45, pp.50-57.

Järvinen, J. and Karjaluoto, H. (2015). The use of Web analytics for digital marketing performance measurement. Industrial Marketing Management, 50, pp.117-127.

Lemon, K. (2016). The Art of Creating Attractive Consumer Experiences at the Right Time: Skills Marketers Will Need to Survive and Thrive. GfK Marketing Intelligence Review, 8(2).

Liben-Nowell, D. and Kleinberg, J. (2007). The link-prediction problem for social networks. Journal of the American Society for Information Science and Technology, 58(7), pp.1019-1031.

Morin, C. (2011). Neuromarketing: The New Science of Consumer Behavior. Society, 48(2), pp.131-135.

Provost, F. and Fawcett, T. (2013). Data Science and its Relationship to Big Data and Data-Driven Decision Making. Big Data, 1(1), pp.51-59.

Rawal, P (2013). AIDA Marketing Communication Model: Stimulating a purchase decision in the minds of the consumers through a linear progression of steps. Irc’s International Journal of Multidisciplinary research in social & management sciences, 1 (1).

Reinartz, W., Thomas, J. and Kumar, V. (2005). Balancing Acquisition and Retention Resources to Maximize Customer Profitability. Journal of Marketing, 69(1), pp.63- 79.

Skiera, B. (2016). Data, Data and Even More Data: Harvesting Insights From the Data Jungle. GfK Marketing Intelligence Review, 8(2).

Tiago, M. and Veríssimo, J. (2014). Digital marketing and social media: Why bother?. Business Horizons, 57(6), pp.703-708.

46

Vargo, S. and Lusch, R. (2004). Evolving to a New Dominant Logic for Marketing. Journal of Marketing, 68(1), pp.1-17.

Wilken, R. (2014). Places nearby: Facebook as a location-based social media platform. New Media & Society, 16(7), pp.1087-1103.

Websites Digitalfans.se. (2018). Digital Fans: Case, This Is Nuts. [online] Available at: http://digitalfans.se/thisisnuts [Accessed 12 Apr. 2018].

Foundr. (2018). Growth Hack Your Facebook Page: We Got 10K Real Likes in 6 Months. [online] Available at: https://foundr.com/growth-hack-your-facebook- page/ [Accessed 20 Apr. 2018].

47

www.kth.se

48