Review of the Tfl Wifi Pilot

Review of the TfL WiFi pilot

Our findings Contents

About Transport for London (TfL) 4 Foreword Part of the Greater London Authority Elizabeth line, modernising Tube services family of organisations led by Mayor of and stations, transforming the road 6 Executive summary London Sadiq Khan, we are the integrated network and making it safer, especially transport authority responsible for for more vulnerable road users, such as 10 Purpose of the pilot delivering the Mayor’s strategy and pedestrians and cyclists. commitments on transport. 14 Customer research We work hard to make journeys easier As a core element in the Mayor’s overall through effective use of technology and 18 Keeping customers informed plan for London, our purpose is to keep data. We provide modern ways to pay London moving, working and growing, through Oyster and contactless payment 20 What data was collected? and to make life in our city better. We cards and provide information in a wide reinvest all of our income to run and range of formats to help people move 22 Protecting customers’ privacy and security improve London’s transport services and around London. 22 Assessing our approach to make it safer, modern and affordable for everyone. We play a central role Real-time travel information is provided 22 Encrypting data collected in delivering the Mayor’s strategy to directly by us and through third party 22 Data storage improve air quality and public health and organisations, which use the data we to make transport accessible to all. make openly and freely available to 24 How successful was the data collection? power apps and other services. Our operational responsibilities include 26 How data can help us improve our services London Underground, London Buses, We listen to, and act upon, feedback 26 Customer information Docklands Light Railway (DLR), London and complaints to constantly improve 38 Operations and safety Overground, TfL Rail, London Trams, our services and work with boroughs, London River Services, London Dial-a- communities, representative groups, 40 Transport planning Ride, Victoria Coach Station, Santander businesses and many other stakeholders 43 Prioritising investment Cycles and the Emirates Air Line. to shape transport provision in London. 45 Conclusion On the roads, we regulate taxis and the Improving and expanding transport in private hire trade, run the Congestion London is central to driving economic Charging scheme, manage the city’s growth, jobs and housing throughout the 580km red route network, operate all United Kingdom. Where possible, we are of the Capital’s 6,300 traffic signals and using our land to provide thousands of work to ensure a safe environment for all new, affordable homes. Our own supply road users. chain creates tens of thousands of jobs and apprenticeships across the country. We are delivering one of the world’s largest programmes of transport capital investment, which is building the

Review of the TfL WiFi pilot 3 Foreword Technology is transforming our lives, from how we work and enjoy our leisure time to the way we travel. For TfL, the dramatic pace of change has presented exciting opportunities.

Electric buses are improving air quality, depersonalised WiFi connection data while state-of-the art signalling can fill in the gaps, as it allows us to means Victoria line trains run every see how people travel beyond the 100 seconds. And smartcard ticketing gateline. It removes the need for costly, is offering customers a wide range of time-consuming surveys and means convenient ways to pay. we can provide detailed customer information for specific times of the The phenomenal rise in the number of day, on individual lines, platforms people using smartphones and mobile and even trains. The potential benefits devices, and the growing appetite for are enormous. information on the go, led us to provide free WiFi at more than 97 per cent of our Given the success of the pilot, we plan stations. We continue to invest in better to collect WiFi data connections across digital customer information and making our network. We will keep customers non-personal data freely and openly informed, respect their privacy and available to app developers. clearly communicate how to opt out.

And there are still untapped In this modern, data-driven world, opportunities that, if explored, can the possibilities are endless. radically improve customers’ journeys. Which route or platform do they use? Our transport system generates huge Do they take the first train or wait for amounts of data. For instance, on a a less crowded one? Do people typical weekday, we collect 19 million choose the fastest route, or the most smartcard ticketing transactions that comfortable? How do customers move offer a valuable insight into the number around our most complex stations? of people using our network. But this Lauren Sager Weinstein doesn’t give us a complete picture – Ahead of and during the pilot, posters kept A four-week pilot in November and Chief Data Officer after they tap in, how do they travel? customers informed December last year revealed that Transport for London

4 Foreword Review of the TfL WiFi pilot 5 Executive summary

Customers are at the heart of all that we do. Every journey matters to us and we work hard to make sure our transport network is safe, reliable and fit for the future. 97% More people than ever are using the of Tube Tube, with more trains running than stations have free WiFi at any time in its history. And we are continuing to transform the way we serve our customers. In addition to the substantial improvements under way on our trains and stations, we are looking to use technology to provide Customer information: could WiFi even better information to help people connection data help us provide better get around London. customer information for journey planning and avoiding congestion? Across the Tube network, 97 per cent of stations now have free WiFi installed, Operations and safety information: which helps people stay up to date could understanding customer while they travel. As a by-product movements in stations help us deploy of providing this service, our router our people to best meet customer network collects information about the needs, and manage disruptions and connections made to the WiFi system. events more effectively? This data could potentially be very useful for understanding travel patterns. Transport planning: by better understanding how our customers From 21 November to 19 December 2016, use the Tube network, could we plan we conducted a pilot to evaluate the timetables, and our station designs usefulness of WiFi connection data. and upgrades, more efficiently? It involved collecting depersonalised data from mobile devices that made Prioritising investment: by measuring connections to our WiFi network at customer footfall and movements 54 stations. We then tested whether through and around stations, could we this could be used to provide benefits assess the effectiveness of our poster in four areas: sites and retail units?

WiFi data can help us understand customers’ travel patterns beyond the gateline

6 Executive summary Review of the TfL WiFi pilot 7 Historically, we have used a combination We relied on guidance from the of data sources to understand customer More than Information Commissioner’s Office movements between and within (ICO), the UK’s independent authority stations. This has included, since 2005, set up to uphold information rights. using depersonalised ticketing data 500 million to look at journey patterns. However, Ahead of the launch of the pilot we depersonalised WiFi connection requests while this data is very useful for VictoriVictoriaa carried out a press and awareness understanding gate-to-gate patterns, Northbound TrainsTrains arriving 90%90% fulfulll campaign. We also pledged to share the it does not tell us which routes Southbound TrainsTrains arriving 75%75% fulfulll results and to communicate our future customers take within stations. • The aggregated data can show plans for WiFi connection data. which sections of our network are Jubilee Because of this, we have supplemented crowded, at what times, and how Jubilee The pilot was successfully delivered, Northbound Trains arriving 90% full our ticketing data with manual customer this changes in response to events Northbound Trains arriving 50% full using in-house expertise, for less than Southbound Trains arriving 75% full surveys. This is time intensive, expensive and network alterations Southbound Trains arriving 25% full £100,000. and limited in detail and reliability. WiFi connection data, on the other • We can use this data to power analytical In view of the clear benefits to us and hand, could provide a better and more tools and services that can improve the our customers, we are now planning to cost-efficient solution. This is what we way we run and plan our network, and formally roll out network-wide WiFi data wanted to test in our pilot. can provide our customers with much collection, so we can better understand more detailed information travel patterns, provide enhanced Over four weeks we collected more than information for our customers, and 500 million depersonalised connection Further details and analysis on all these improve our planning and operations. requests, and our technical process ran points are outlined throughout this report. without error. Having collected the data, We will continue to inform our we needed to create new analytic tools Openness and transparency is customers about any WiFi connection and algorithms that could improve our fundamental to the way we operate. collection, the purpose behind it, and understanding of travel patterns. From the first time we looked at how we will use the data. And we will whether we could use WiFi connection make sure people know how to opt out. From our data analysis, we have been data to understand travel patterns WiFi connection data could help us improve customer information for journey able to conclude that: on our network, we recognised planning and avoiding congestion the importance of making sure our • WiFi data can help us understand customers clearly understood what we the paths customers take in stations, were proposing and why. We started the platforms and lines they with customer research and focus use, which route they take when groups to test our proposal. Next, we many options exist and where assessed it through a structured Data they interchange Protection Impact Assessment.

8 Executive summary Review of the TfL WiFi pilot 9 Purpose of the pilot

Harrow- Hendon Central With 45 per cent of the Tube network Figure 1: Map of the pilot area Preston Highgate Tottenham Blackhorse on-the-Hill Road Seven Kingsbury Sisters Hale Road Rayners Lane Manor House Walthamstow Redbridge in tunnels and stations as deep as 58.5 Brent Cross Archway West Northwick Gospel Central Harrow Park Oak Wanstead Gants Golders Green Upper Holloway Stamford Hill metres, there are many areas where South Kenton Hampstead Hill Neasden Heath Walthamstow Leytonstone South Harrow Wembley Park Hampstead Queen’s Road Leyton North Wembley Tufnell Park Arsenal Midland Road cellular networks do not provide Dollis Hill Finsbury Stoke Newington Wembley Central Kentish Park Willesden Green Holloway Road St. James Street Leytonstone Wanstead connectivity. In partnership with Virgin Finchley Road Town West Kentish Town High Road Park Stonebridge Park & Frognal Belsize Park Rectory Road Forest Sudbury Hill Kilburn Caledonian Road Clapton Leyton Gate Media, we started installing WiFi in Harlesden Highbury & Dalston Kensal Brondesbury West Chalk Farm Camden Islington Kingsland Rise Park Hampstead Stratford Willesden Junction Road Hackney International Maryland Caledonian Hackney stations in 2012, and it is now available Camden Town Downs Brondesbury Road & Central Sudbury Town Finchley Road Canonbury Kensal Green Barnsbury Stratford at 97 per cent of all Tube stations. This Swiss Cottage Mornington Dalston Junction Homerton Hackney Queen’s Park Kilburn South St. John’s Wood Crescent Wick High Road Hampstead enables millions of customers to get King’s Cross London Fields Stratford Alperton St. Pancras Haggerston Greenford High Street Kilburn Park Paddington Marylebone Cambridge Heath online and find their way around using Edgware Road Baker Great Portland Euston Street Maida Vale Street Hoxton Bethnal Green Pudding Abbey Warwick Angel Mill Lane Road Upton up-to-the-minute travel information. Perivale Avenue Bethnal Green Mile End Park Euston Old Street Bow Warren Street Edgware Square Road Plaistow Road Royal Oak Farringdon Shoreditch Hanger Lane Regent’s Park Russell High Street Stepney Green Bromley- West Ham Connectivity data is generated as a Westbourne Park by-Bow Square Bow Barbican Park Royal Bayswater Goodge Church by-product of providing this service. Ladbroke Grove Street Aldgate Bond Oxford East Whitechapel Devons Road We already use a range of data, such as Latimer Road Street Circus Moorgate Liverpool Star Lane North Ealing Marble Arch Street Tottenham Holborn Chancery Lane Langdon Park East White Shepherd’s Notting Court Road Aldgate aggregated and depersonalised Oyster Acton City Bush Hill Gate Bank Covent Garden All Saints Ealing Broadway Lancaster Canning and contactless payment transactions, West North Holland Queensway Gate St. Paul’s Town Acton Acton Park Green Park Leicester Square Limehouse Poplar Wood Lane Hyde Park Corner to understand how customers travel Piccadilly Cannon Street Shadwell Westferry Blackwall East Acton Central Shepherd’s High Street Kensington Circus Monument Tower India Ealing Common Bush Market Kensington Hill across London. While this tells us (Olympia) Mansion House Fenchurch Street Tower South Acton Knightsbridge Charing Gateway Wapping West India Goldhawk Road Cross Blackfriars Quay where customers enter and exit the South Barons Gloucester River Thames Ealing Court Road St. James’s Acton Town Hammersmith Rotherhithe Tube network, it does not reveal Victoria Park Temple Canary Wharf Emirates Chiswick Turnham Stamford Ravenscourt West Sloane Westminster Embankment London Bermondsey Canada North Northﬁelds Earl’s South Greenwich Greenwich anything about their route choices, Park Green Brook Park Kensington Court Kensington Square Bridge Water Heron Quays Peninsula Boston Manor such as which platforms and lines Gunnersbury Waterloo South Quay Osterley West Brompton Crossharbour they used, where they interchanged, Hounslow East Southwark Surrey Quays the paths taken to move through Hounslow Central Mudchute Kew Gardens Pimlico Borough and around stations, and how they Fulham Broadway Lambeth North Island Gardens Richmond Parsons Green Imperial Wharf responded to network changes. Putney Bridge Cutty Sark for Maritime Greenwich Queens Road Peckham Greenwich East Putney Vauxhall New Elephant & Castle Cross Gate Owing to the layout of the Tube New Cross Southﬁelds Deptford Bridge Brockley network, there are many route Peckham Rye Elverson Road Wimbledon Park Kennington Wandsworth Clapham Oval Honor Oak Park options that customers could take. Wimbledon Junction Road Lewisham Stockwell To understand these choices and help Clapham High Street Denmark Hill Forest Hill Clapham North Sydenham us plan and operate the network, Clapham Common we have historically relied on paper Stations included in the pilot: Dundonald Road Brixton surveys and manual counts, which Aldgate Chalk Farm Green Park London Bridge Piccadilly Circus Tower Hill are expensive, time consuming and Angel Chancery Lane Holborn Mansion House Regent's Park Tufnell Park limited in detail and reliability. Baker Street Charing Cross Kennington Monument Russell Square Victoria Bank Covent Garden Kentish Town Moorgate St. James's Park Warren Street Belsize Park Dollis Hill Kilburn Mornington Crescent St. Paul's Waterloo Blackfriars Elephant & Castle King's Cross St. Pancras Neasden St. John's Wood Wembley Park Borough Embankment Lambeth North Old Street Stockwell West Hampstead Camden Town Euston Leicester Square Oval Swiss Cottage Westminster Cannon Street Finchley Road Liverpool Street Oxford Circus Temple Willesden Green

10 Purpose of the pilot Review of the TfL WiFi pilot 11 WiFi connection data could provide The pilot was carried out at 54 a greater understanding of crowding London Underground stations within and collective travel patterns across Zones 1-4. Figure 1 on p10-11 shows the network, while also being more the area covered. cost efficient. Individuals who did not want us to We conducted this pilot to collect collect the connections made by their depersonalised WiFi data and analyse device were advised to either turn off whether it could help us provide four WiFi, or put their device into airplane important benefits: mode while at one of the stations included in the pilot. Customer information: could WiFi connection data help us improve All TfL mobile devices issued to customer information for journey employees were excluded to planning and avoiding congestion? avoid distorting the information being collected. Operations and safety information: could understanding customer movements in stations help us deploy our people to best meet customer needs, and manage disruptions and events more effectively?

Transport planning: by better understanding how our customers use the Tube network, could we plan timetables, and our station designs and upgrades, more efficiently?

Prioritising investment: by measuring customer footfall and movements through and around stations, could we assess the effectiveness of our poster sites and retail units?

Fifty-four Tube stations were included in the pilot

12 Purpose of the pilot Review of the TfL WiFi pilot 13 Customer research

From the point we first considered using We explored a number of options for WiFi connection data to identify travel using data generated by mobile devices. patterns, we recognised the importance These are shown in Figure 2 on p16-17. of making sure our customers clearly The overall feedback was positive for understood what we were proposing all scenarios, particularly when our and why. To gauge people's views, we approach for collecting the data was carried out customer research and set transparent and offered clear and up focus groups. This tested a number tangible customer benefits, with the of uses for mobile phone data to: ability to opt out.

• Understand assumptions around Collecting WiFi data on the Tube was the data we currently have and better received than other scenarios what we do with it as it had a clear purpose and benefits. These findings helped us shape our pilot • Explore attitudes towards the different and highlighted that we needed to: types of mobile connection data, both in general and distinguishing • Use a number of communication between types (Bluetooth, cellular, methods to inform customers of WiFi and GPS) the pilot's purpose

• Understand how customers feel • Communicate clearly that data would about this data being used to improve not be used to identify individuals journeys for them personally and across the system as a whole • Use the data to offer clear customer benefits • Guide how we could communicate our plans for using such data

• Help us understand the situations in which using such data was considered acceptable

During our research, customers recognised the benefits of collecting WiFi data on the Tube

14 Customer research Review of the TfL WiFi pilot 15 Caution Caution Caution Caution Caution Caution Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Figure 2: Scenarios presented during the focus groups 96601 96601

Caution Caution Caution Caution Caution Caution Sliding doors Sliding doors Scenario 1: Road beacons – for real-time traffic flow and congestion information Scenario 3: WiFi connection data onSliding doors the Tube – forSliding doors better customer Sliding doorinformations Sliding doors

96601 96601

Trains arriving every 2 mins Train crowding levels on arrival at this station

1. Next train 1 min Victoria 2. Next train 3 mins Northbound Trains arriving 90% full Southbound Trains arriving 75% full 10 minutes 3. Next train 5 min Jubilee Victoria Northbound Trains arriving 50% full Northbound Trains arriving 90% full Southbound Trains arriving 25% full Southbound Trains arriving 75% full

Jubilee Northbound Trains arriving 50% full Southbound Trains arriving 25% full Trains arriving every 2 mins Train crowding levels on arrival at this station

Caution Caution Caution Caution Caution Caution Sliding doors Sliding doors Sliding doors Sliding doors 1. NextSliding doors train 1 min Sliding doors Victoria 2. Next train 3 mins Northbound Trains arriving 90% full 3. Next train 5 min Southbound Trains arriving 75% full

96601 96601 Jubilee Victoria Northbound Trains arriving 50% full Northbound Trains arriving 90% full Southbound Trains arriving 25% full Southbound Trains arriving 75% full Jubilee Northbound Trains arriving 50% full Southbound Trains arriving 25% full

38 REGISTER YOUR DEVICE REGISTER YOUR DEVICE

Scenario 2: Mobile phones – to understand movement and support planning decisions Scenario 4: Location data – for up-to-date information and personalised alerts Trains arriving every 2 mins Train crowding levels on arrival at this station

1. Next train 1 min Victoria 38 2. Next train 3 mins Northbound Trains arriving 90% full 3. Next train 5 min REGISTERSouthbound YOURTrains arrivingDEVICE 75% full REGISTER YOUR DEVICE Piccadilly line has been suspended Jubilee Dear Mr Brown, Dear Mrs Johnston, between Green Park Northbound Trains arriving 50% full TfL travel alert Victoria Your regular docking TfL travel alert and Earl’s Court. We noticed that you station on Pall Mall A2 roadworks have didn’t tap out at Northbound Trains arriving 90% full Southbound Trains arriving 25% full Delay on number 8 Wednesday, 14:39 is currently empty been completed early Morden yesterday Southbound Trains arriving 75% full bus route between but a bike delivery is – all lanes will be morning, so we have St. Pauls and Holborn. due at 11am. Sorry for reopened from 10am auto-completed your REGISTER YOUR DEVICE 38 REGISTER any inconvenience. journey to avoid you Jubilee being charged the Yours sincerely, YOUR DEVICE Northbound Trains arriving 50% full maximum fare. Southbound Trains arriving 25% full

REGISTER YOUR DEVICE REGISTEREmail Push notiﬁcation Text message Push notiﬁcation Email 38 YOUR DEVICE REGISTER YOUR DEVICE REGISTER YOUR DEVICE Piccadilly line has Dear Mr Brown, been suspended Dear Mrs Johnston, between Green Park TfL travel alert Your regular docking TfL travel alert We noticed that you and Earl’s Court. station on Pall Mall A2 roadworks have didn’t tap out at Delay on number 8 Wednesday, 14:39 is currently empty been completed early Morden yesterday bus route between but a bike delivery is – all lanes will be morning, so we have St. Pauls and Holborn. due at 11am. Sorry for reopened from 10am auto-completed your 38 any inconvenience. journey to avoid you being charged the Yours sincerely, maximum fare.

Email Push notiﬁcation Text message Push notiﬁcation Email

Piccadilly line has Dear Mr Brown, been suspended Dear Mrs Johnston, between Green Park TfL travel alert Your regular docking TfL travel alert and Earl’s Court. We noticed that you station on Pall Mall A2 roadworks have didn’t tap out at Piccadilly line has Delay on number 8 Wednesday, 14:39 is currently empty been completed early Morden yesterday bus route between Dear Mr Brown, been suspended Dear Mrs Johnston, but a bike delivery is – all lanes will be morning, so we have St. Pauls andEmai Holborn.l Push notification Text message Push notification Email between Green Park TfL travel alert due at 11am. Sorry for reopened from 10am auto-completed your Your regular docking TfL travel alert We noticed that you any inconvenience.38 journey to avoid you and Earl’s Court. being charged the station on Pall Mall A2 roadworks have didn’t tap out at Yours sincerely, Delay on number 8 Wednesday, 14:39 maximum fare. is currently empty been completed early Morden yesterday bus route between but a bike delivery is – all lanes will be morning, so we have St. Pauls and Holborn. due at 11am. Sorry for reopened from 10am auto-completed your any inconvenience. journey to avoid you being charged the Yours sincerely, maximum fare. Email Push notification Text message Push notification Email

Email Push notiﬁcation Text message Push notiﬁcation Email Keeping customers informed

We are committed to being open and transparent about how we use data. The More than ICO is the independent regulator of Save the data for transport trial The bigger picture Data collection Shashi Verma, chief technology pilot launches in officer at TfL, said: ‘This short trial Opportunity personal data in the UK. We considered will help us understand whether WiFi Tube stations. connection data could help us plan to opt out and operate our transport network FROM today, passengers’ WiFi more effectively for customers. Posters will be displayed in connection data will be collected at ‘Historically, if we wanted to know stations to let people know 54 stations within Zones 1-4 to give TfL how people travelled we would have the trial is taking place. the ICO’s guidance on WiFi location a better understanding of how people to rely on paper surveys and manual If any customers wish to opt use the services. counting, which is expensive, time out of the trial, they should turn WiFi connection requests will be picked consuming and limited in detail off their WiFi while passing up from mobile devices as passengers and reliability.’ through the station. pass through stations. When a device is Sue Daley, head of big data, cloud & All data will be automatically posters WiFi-enabled, it searches for a WiFi mobile at techUK, said: ‘TfL is unlocking de-personalised and no data will be analytics and its code of practice on network by sending out a unique the power of data to gain insights into made available to any third-parties. identifier to nearby routers. how passengers are using the network 300 The data will give a more detailed and drive its transformation into a understanding of how people move smart transport system. By understanding how people through stations, change between ‘The availability of big data analytics move through and around stations, across the pilot area services and how congestion tools and technologies means that there is a possibility of increasing privacy notices, transparency and control. occurs. This should help TfL improve organisations, of all sizes and sectors, revenue from companies who services and travel information, and are increasingly able to make data advertise on poster sites or rent retail help investment. driven decisions that can make a real units to reinvest into improving It will also allow TfL to manage difference to customers’ lives. In this services across London. disruptions and events better, and case, it will mean more accurate improve planning of timetables and passenger insights and easier journeys ■ To find out more about the trial, station design. for customers.’ visit tfl.gov.uk/privacy Consistent with best practice, we Londoners encouraged adopted the ICO’s ‘layered approach’. to sign up to RideLondon THE ballot for a place in next year’s Prudential RideLondon-Surrey 100 is now open. The ride, which takes place on This involved using a number of Sunday July 30, offers cyclists the chance to experience some of London’s most historic streets, free from cars, buses and other vehicles. Starting at Queen Elizabeth Olympic Park, the route heads out through communication methods to make central London and into Surrey, where riders will pedal through picturesque villages. After taking on the challenges of the Leith Hill and Box Hill climbs, cyclists will return to London for an STATION epic finish on The Mall in front of sure customers knew that we would Buckingham Palace. The event has raised millions of pounds for charity in the four years since it launched. The ballot is open until 5pm on January 6 or when 80,000 be collecting data and understood the applications have been received – ■ To enter the ballot, log on to www.PrudentialRideLondon.co.uk and whichever is sooner. follow the entry details benefits of doing so. For more articles and to keep up to date with TfL announcements, visit tfl.gov.uk/news London travel advice: plan your journey now

London Travelwatch London’s transport watchdog call 020 3176 2999, or visit The week before launching the pilot, www.londontravelwatch.org.uk 1 we issued a press release that set out * Service and network charges apply. See tfl.gov.uk/terms for details.

Newspapers left on the Tube can jam doors and cause delays to your journey. the benefits and scope of the project. Take your newspaper with you or put it in a recycling bin. The views expressed are those of TfL only and are not those of Metro. This resulted in the pilot being widely reported in the media.

Caution More than 300 large posters were put Sliding doors We publicised details on our daily travel up across the pilot area, including on page in Metro on 21 November (pictured platforms and at station entrances. They right). The newspaper distributes more directed customers to our webpage for than 764,000 copies a day and is read by more information. in excess of 1.9 million people. We used social media to inform people, Throughout the pilot, a new webpage was including our 1.6 million Twitter followers available on tfl.gov.uk/privacy with more and 386,000 followers of the TfL information and answers to anticipated Facebook page. customer questions. Following feedback that this information was not prominent Employees at stations and our Customer enough for mobile users who needed Contact Centre received briefing packs to to scroll down, we made changes and help them answer customer questions. displayed a link at the top of the privacy We published an article, similar to the page for easier navigation. press release, on our developer blog. Important stakeholders such as the Greater London Authority, ICO and trade 1. tfl.gov.uk/info-for/media/press-releases/2016/ unions were informed and offered more november/wifi-trial-to-help-give-customers- detailed briefings if required. better-journeys

18 Keeping customers informed Review of the TfL WiFi pilot 19 What data was collected?

When a mobile device such as a During the pilot, if a device was near smartphone or tablet has WiFi enabled, one of the 1,070 WiFi access points in More than it will search for a WiFi network to the designated area, and it had WiFi connect to. This involves the device enabled, we would have collected the sending out a probing request that request(s) to connect, even if the device 509 contains an identifier specific to that did not subsequently do so. Once the device, known as a Media Access Control device had connected, we also gathered million (MAC) address. If a WiFi network is these requests and authentications. probing requests

Caution Caution Caution Caution Caution Caution found that is known to the device, it will No web browsing data or data from were collected at Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors automatically connect. If the device finds website cookies was gathered. 96601 96601 unknown networks, it lists these in the device settings so the user can decide For customers who did not have WiFi 54 which, if any, to connect to. enabled, no data was collected. stations from 5.6 Figure 3: A fictional example of the data collected million devices Hashed MAC address Date Time Access point x1Jx7F893lL4jO 01/12/2016 07:12:22 London Bridge 01 x1Jx7F893lL4jO 01/12/2016 07:13:22 London Bridge 15 x1Jx7F893lL4jO 01/12/2016 07:14:06 London Bridge 10 x1Jx7F893lL4jO 01/12/2016 07:17:04 Bank 10 Throughout the pilot we only A fictional example of the type of data x1Jx7F893lL4jO 01/12/2016 07:17:26 Bank 30 collected the minimum information collected is outlined in Figure 3. It shows x1Jx7F893lL4jO 01/12/2016 07:20:26 Old Street 06 required to test whether WiFi data only the MAC address, date, time and could be used to provide the customer access point for a device travelling from x1Jx7F893lL4jO 01/12/2016 07:20:50 Old Street 02 benefits we had identified. This included London Bridge, through stations on the x1Jx7F893lL4jO 01/12/2016 07:22:52 Angel 06 an encrypted, depersonalised version Northern line (Bank branch) to Belsize x1Jx7F893lL4jO 01/12/2016 07:25:18 King’s Cross St. Pancras 22 of the device MAC address, the date Park. As we know where each access x1Jx7F893lL4jO 01/12/2016 07:26:01 King’s Cross St. Pancras 18 and time the device broadcast its MAC point is located (platform, ticket hall, etc), x1Jx7F893lL4jO 01/12/2016 07:27:43 Euston 14 address, the access point it connected we can understand where in the station to, the device manufacturer and the the device was when it connected. x1Jx7F893lL4jO 01/12/2016 07:30:19 Camden 06 device association type. x1Jx7F893lL4jO 01/12/2016 07:32:48 Chalk Farm 02 x1Jx7F893lL4jO 01/12/2016 07:33:08 Chalk Farm 04 x1Jx7F893lL4jO 01/12/2016 07:34:47 Belsize Park 04

20 What data was collected? Review of the TfL WiFi pilot 21 Protecting customers’ privacy and security

Protecting the privacy and security of In line with the ICO’s WiFi Analytics our customers’ data is of paramount Guidance3, our irreversible encryption importance and we recognise our method included a ‘salt’. This further responsibilities as a custodian of the enhanced security by adding an unknown personal data of millions of people. character string to each MAC address We understood that recording the before encryption. Even if a MAC address location of a customer’s device MAC and the encryption mechanism is known, address at a specific place and time it is impossible for the original MAC could be considered as personal data. address to be identified without knowing We employed a number of controls to the salt. This reduces both the possibility make sure the pilot fully complied with of a single known MAC being identified the Data Protection Act 1998. We also in the data, and a look-up table of all briefed the ICO about our plans and met possible MAC addresses being created representatives to discuss these further. and joined to the data.

Assessing our approach The salt is not known by any individual To ensure our communications and and was destroyed on the day the data data processing approach was collection ended. Therefore, we consider appropriate, we completed an extensive the data to be anonymous and are Data Protection Impact Assessment unable to identify any specific device. before the pilot took place. It followed As we cannot process known MAC our formal TfL governance and project addresses in the same manner as we management processes. did in the pilot, we are unable to complete any Subject Access Request4 Encrypting data collected for the data we collected. Each MAC address collected was automatically depersonalised Data storage (pseudonymised2) and encrypted to The pilot data was stored in a restricted prevent the identification of the original area of a secure server and has not MAC address and associated device. been linked to any other customer information. Access was restricted to a small team that has regular privacy and data protection training. The data collected has not been, and never will 2. The process of distinguishing individuals in a be, shared with any third parties. dataset by using a unique identifier that does not reveal their ‘real world’ identity, in accordance with the ICO’s Anonymisation Code of Practice 3. www.ico.org.uk/for-organisations/guidance- index/data-protection-and-privacy-and- electronic-communications 4. tfl.gov.uk/corporate/privacy-and-cookies/ WiFi connection data collected during access-your-data the pilot will remain anonymous

22 Protecting customers' privacy and security Review of the TfL WiFi pilot 23 How successful was the data collection?

From the 54 stations included in the Using this approach, we constructed Figure 4: WiFi probing requests collected each day pilot, we collected 509 million probing 42 million journeys from five million requests from 5.6 million devices. devices during the pilot. For some King’s Cross St. Pancras generated the devices, we either collected single 25,000,000 most, with 37.6 million. The fewest probing requests or the MAC address were observed at Dollis Hill, where was randomised so we were unable to 10,000 were collected from seven constuct journeys. access points. 20,000,000 The next step was to break these Figure 4 shows the number of WiFi journeys down into 'movement types'. connections automatically collected Using advanced analytic techniques, on each day of the pilot. On average, we classified these as: 15,000,000 more probing requests were generated on weekdays, when most people travel. • Entry or exit – where a device Thursday 15 December saw the highest was entering or exiting the Tube number of WiFi connections, with 21.2 network at a specific point 10,000,000 million probing requests collected. • Pass-through – where a device was It typically took between 20 and 40 seen in a station as they passed minutes from a device connecting to through while on a train 5,000,000 our network to collect the data in our secure server. Across the entire • Interchange – where a device was pilot period, the longest lag was 102 changing from one line to another minutes between device connection within a station and collection. We believe we can 0 21 22 23 24 25 26 27 28 29 30 1234 5678910111213141516171819 achieve real-time data collection based • Sub-categories of movements – where on what we learnt from the pilot. we determined that a device was November 2016 December 2016 boarding or alighting a train service Transforming WiFi data into journeys To translate and interpret the data, we The reason for creating this Data collected on weekdays (Monday to Friday) applied ‘big data’ techniques to convert classification system was not to track Data collected on weekends (Saturday and Sunday) these individual connections into a individuals or pinpoint the actions made structured format. by specific devices, but to understand, in total, the types of movements taken at Firstly, we linked individual connections different locations on the network. from a device so that we could create ‘journeys’ – an end-to-end trip that is comparable to, but more detailed than, what we get from our ticketing data.

24 How successful was the data collection? Review of the TfL WiFi pilot 25 How data can help us improve our services

We carried out the pilot to test whether The following insights from the pilot Figure 5: Busiest times at Oxford Circus station, using ticketing and WiFi data WiFi data could provide four potential illustrate how this information may benefits. Analysis has demonstrated help customers improve the most that it would add considerable value relevant aspects of their journey: and enable us to plan and operate our transport services more effectively. This • When are individual stations, includes providing better information platforms and trains busiest? and more detailed and timely updates on congestion so customers can choose • How busy are the different line and travel options that best suit their route options? circumstances. Our findings for each of the four potential benefits outlined on • How does disruption affect my journey? p13 are positive and these are explored in more detail below. Which times are busiest? WiFi connection data is an accurate Customer information and reliable way to quantify crowding WiFi data allows us to measure and at a station at different times of the understand crowding and demand on day. Unlike ticketing data, which only our network in greater detail than was shows customers entering and exiting previously possible. We can determine stations, WiFi captures interchanges. the paths customers take within and This is especially important when between stations. This, in turn, enables measuring crowding levels in large us to identify crowding levels in a stations where many customers station, on a specific platform, on a change between services. line or on a train at a given time of day. Figure 5 highlights the demand at Oxford 04:30 07:00 09:30 12:00 14:30 17:00 19:30 22:00 Taken together, we can use this Circus station on a typical weekday information to tell people about during the pilot. The demand profile network conditions at every stage from Oyster ticketing data is shown in WiFi (%) of their journey. This can help our blue, and demand estimated from WiFi customers to plan routes that best suit connections is orange. The differences Oyster (%) them, based on the network conditions can be attributed to Oxford Circus being at the time they are travelling. a major interchange station where the Bakerloo, Central and Victoria lines meet. Ticketing data cannot show the sharp increase in crowding between 08:30 and 09:00 because it only captures customers entering and exiting stations.

26 How data can help us improve our services Review of the TfL WiFi pilot 27 WiFi can also accurately capture To understand where crowding occurs crowding even when, on occasion, on each line, we calculate the number gates are left open, for example of people on each train based on where during large events or disruption. and when devices connect to the WiFi The information we currently give to network, in combination with train customers is based on ticketing data locations. Traditionally, this information and surveys for a typical day. WiFi has only been available for 15-minute would enable us to provide more intervals and is collected when we survey accurate travel information on a our customers on non-disrupted days. continuous basis, under changing network conditions. However, this is relatively expensive and does not capture times when the This data would be particularly useful network is disrupted. WiFi data will allow to customers who would rather avoid us to provide continuous, responsive crowding, even if it meant increasing estimates of demand on specific their travel time. It would allow services, enabling passengers to them to re-time and/or re-route their make informed decisions about journeys to avoid the busiest sections the journeys they make. of the network.

How busy are the different line and route options? We can use WiFi data to understand where and when crowding occurs on trains and on specific sections of the Tube network. This information can tell our customers how busy different route options are for the same journeys.

Data collected from WiFi connections will allow us to provide more detailed customer information

28 How data can help us improve our services Review of the TfL WiFi pilot 29 JOURNEY RESULTS

From: Camden Town Underground Station Travel preferences & accessibility

To: St. James’s Park Underground Station Showing the fastest routes Using all transport modes Max walk time 40 mins Leaving: Monday, 13:10

A B Edit journey Add favourites Edit preferences ˅

Fastest by public transport

For example, someone planning to Many customers want to understand available through our travel tools. It 2 mins Northernthe Victoria line to Embankment line Underground between Statio nEuston 6 mins Victoria line to Victoria Underground Station travel between Camden Town and how busy the train(s) and interchange highlights theDeparts averageArrives demand across and Green Park. 1 min St. James’s Park is presented with station(s) could be so they know if they each section13:07 of these13:22 two journeys15 mins District line or Circle line to St. James’s Park Underground Station i This journey has additional information two options when using TfL Journey are able to board the first train. Data during the morning peakView Details period This crowding can be avoided by taking Planner, as shown in Figure 6. gathered from WiFi allows us to better (between 7:00 and 10:00). Assuming the Northern line (Charing Cross branch)

advise customers by providing this customers can board the very crowded10 mins Northerndirect lin eto to Embankment Embankment. Underground Statio Somen customers

These options, however, do not help level of detail. train in the first option, this is quicker 2 mins Districtmay lin alwayse to St. James’ prefers Park Underground the quicker Station of the them make their decision based on although significantlyDeparts Arrives more crowded, two options, but by providing these 13:10 13:27 17 mins i This journey has additional information factors other than estimated time Figure 7 shows how we could use WiFi especially on the branch between details we can help them make View Details and the number of interchanges. data to improve the route information Camden Town and Euston, and on informed decisions.

Figure 6: Journey planning options for a trip between Camden Town and St. James’s Park Figure 7: Enhanced travel tools using crowding information

JOURNEY RESULTS JOURNEY RESULTS

From: Camden Town Underground Station Travel preferences & accessibility From: Camden Town Underground Station Travel preferences & Customer journey options accessibility To: St. James’s Park Underground Station Showing the fastest routes Using all transport modes To: St. James’s Park Underground Station Max walk time 40 mins Showing the fastest routes Seats free Leaving: Monday, 13:10 Leaving: Monday, 13:10 Using all transport modes Camden Town Some standing A Max walk time 40 mins A B Edit journey Add favourites Edit preferences ˅ B Edit journey Add favourites Busy Mornington Edit preferences ˅ Crowded Crescent Very crowded Full

Fastest by public transport Fastest by public transport Euston

Warren Euston 2 mins Northern line to Embankment Underground Station 2 mins Northern line to Embankment Street Square Underground Station 6 mins Victoria line to Victoria Underground Station Goodge Departs Arrives Departs Arrives 6 mins Victoria line to Victoria 1 min Oxford Circus Street 13:07 13:22 15 mins District line or Circle line to St. James’s Park Underground Station 13:07 13:22 15 mins Underground Station 1 mins District line or Circle line to i This journey has additional information View Details View Details St. James’s Park Underground Station Green Park Leicester 10 mins Northern line to Embankment Square 10 mins Northern line to Embankment Underground Station Underground Station Charing Departs Arrives 2 mins District line to St. James’s 2 mins District line to St. James’s Park Underground Station Cross Departs Arrives Park Underground Station Victoria 13:10 13:27 17 mins 13:10 13:27 17 mins i This journey has additional information i This journey has Embankment View Details additional information St. James’s Park View Details

30 How data can help us improve our services Review of the TfL WiFi pilot 31

JOURNEY RESULTS

From: Camden Town Underground Station Travel preferences & Customer journey options accessibility To: St. James’s Park Underground Station Showing the fastest routes Seats free Leaving: Monday, 13:10 Using all transport modes Camden Town Some standing A Max walk time 40 mins B Edit journey Add favourites Busy Mornington Edit preferences ˅ Crowded Crescent Very crowded Full

Fastest by public transport Euston

Warren Euston 2 mins Northern line to Embankment Street Square Underground Station Goodge Departs Arrives 6 mins Victoria line to Victoria Oxford Circus Street 13:07 13:22 15 mins Underground Station 1 mins District line or Circle line to View Details St. James’s Park Underground Station Green Park Leicester 10 mins Northern line to Embankment Square Underground Station Charing Departs Arrives 2 mins to District line St. James’s Cross 13:10 13:27 17 mins Park Underground Station Victoria i This journey has Embankment View Details additional information St. James’s Park In addition to being able to provide an We could use this information to tell Figure 8: Estimated number of passengers on southbound Victoria line trains insight into network-wide crowding customers how their journeys may be departing from Euston, on 9 December 2016 levels, WiFi connection data could affected by crowding or congestion enable us to offer crowding information on the network. It could be presented

Caution Caution Caution Caution Caution Caution for individual trains. at station and line level, allowing 7:50-7:55 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 100+ passengers people to make decisions about the While train-level demand remains route they take. We currently provide

Caution Caution Caution Caution Caution Caution relatively consistent across peak periods, information on the busiest times, but 7:55-8:00 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 300+ passengers there are some fluctuations in crowding WiFi connection data would allow us levels between neighbouring trains. to provide greater detail.

Caution Caution Caution Caution Caution Caution This is partly driven by the number of 8:00-8:05 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 500+ passengers passengers on board when the train How long will my transfer take? arrives, the amount of people attempting We can use WiFi data to understand the

Caution Caution Caution Caution Caution Caution to board, and the number alighting paths customers take within stations, and 8:05-8:10 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 900+ passengers from previous services who may still be how long it takes to move around the working their way along the platform. station under different conditions. This

Caution Caution Caution Caution Caution Caution gives us a detailed picture of the busiest 8:10-8:15 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 700+ passengers If we can provide information for areas at different times of the day. crowding levels on specific trains,

Caution Caution Caution Caution Caution Caution customers may choose to wait for the Figure 9 on p34 shows how customers 8:15-8:20 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 900+ passengers next one if it means a more comfortable move from the Northern line (Bank journey. This would also improve the branch) northbound platform to the

Caution Caution Caution Caution Caution Caution service for everyone by smoothing Victoria line southbound platform at 8:20-8:25 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Over 1,100 passengers demand over the peak period. Euston station. Most people use the shortest path through the passageway.

Caution Caution Caution Caution Caution Caution Figure 8 shows the estimated number of However, about a third head up to the 8:25-8:30 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 900+ passengers passengers on southbound Victoria line main concourse then go down to the services departing from Euston during Victoria line platform. Five per cent take

Caution Caution Caution Caution Caution Caution the busiest times on the morning of more complex paths. Those who use the 8:30-8:35 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 700+ passengers 9 December 2016. The busiest services passageway take around one to three carry more than 1,100 people. Between minutes to travel between the platforms

Caution Caution Caution Caution Caution Caution 08:20 and 08:25, three trains departed and those who use the concourse take 8:35-8:40 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 500+ passengers with more than 1,000 customers, resulting around three to five minutes. This could in very crowded conditions. By contrast, be useful to less frequent travellers who

Caution Caution Caution Caution Caution Caution trains departing in the preceding and want to understand transfer times and 8:40-8:45 Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors Sliding doors 700+ passengers following five minutes were around 10 per levels of crowding in stations. cent quieter, which represents around 15 fewer people per carriage. Not crowded Extremely busy

32 How data can help us improve our services Review of the TfL WiFi pilot 33 Figure 9: How customers move from the Northern line (Bank branch) northbound platform to the Victoria line southbound platform at Euston station 68% of customers use the shortest route at Euston station

Northern line (southbound)

(southbound) Customers choose different routes through Victoria line stations, depending on their needs

When choosing a route through a station, How does disruption affect my journey? customers have different requirements Another potential benefit is the Victoria line (nor and preferences. Detailed information responsiveness of the data to real-time 3-5 thbound) about walk times and crowding can help events. Currently, a great deal of our people make the best decision for them. understanding of crowding is based on For example, someone with specific surveys that only show a snapshot of minutes accessibility requirements, or passengers the network when it is operating without Northern line (nor travelling with buggies or luggage, would disruption. We then make assumptions 1-3 thbound) benefit from knowing the least crowded about likely customer responses, based path from one platform to another. on their typical behaviour. minutes

34 How data can help us improve our services Review of the TfL WiFi pilot 35 Figure 10: Disruption that resulted from the Waterloo & City line suspension With WiFi data collection, which transfer on to the Circle and District is continuous and offers a larger lines, which saw an additional 6,000 sample, we can better understand customers travelling east from Waterloo & City line suspended, which led to customer reactions. Embankment towards Monument. 08:00-09:00 customers making a number of different travel choices to get to Bank station On the morning of 9 December 2016, The impact of these various responses the Waterloo & City line was suspended meant customers on the Northern, between 08:00 and 09:00 owing to a Bakerloo, and Jubilee lines from signal failure. Under normal network Waterloo experienced busier trains and conditions we would expect the line interchanges than usual. Embankment Bank between Waterloo and Bank to be one station was particularly affected, with of the most crowded sections on the many more people changing here Leicester Square entire network. from the Northern and Bakerloo lines Charing Cross than normal – approximately 150 extra With the line unavailable, we saw passengers for each train arriving – and Victoria customers make a number of different many more attempting to board the travel choices to complete their journey. Circle and District lines. London Bridge Around 4,000 decided to use the Jubilee Waterloo Higher demand line to London Bridge, then take the This data is valuable from both a Unchanged Northern line to Bank. customer and operational perspective Waterloo Suspended Waterloo & City line as it enables a more informed response In that hour alone, 3,000 more customers to changes in network conditions. It may travelled from Waterloo on the Bakerloo help customers plan their journeys and Embankment line to Embankment, then continued on allow us to look at the advice we give Waterloo the Circle or District line to Monument. them, how and where we support them Monument/Bank We also saw significant additional as they travel across the network, where volumes of customers go north on the we deploy our staff to offer assistance Northern line to Embankment and north and how we operate our stations and Embankment on the Jubilee line to Westminster. trains to ensure the quickest and most Waterloo All three of these responses meant a comfortable alternative journeys. Westminster Monument/Bank

Waterloo Westminster

London Bank Bridge Waterloo

Additional trains full of people using the route London Bank Bridge Waterloo 36 How data can help us improve our services Review of the TfL WiFi pilot 37 Operations and safety was likely to have been caused by an Figure 12: The change in Euston platform activity prior to the closure WiFi data could prove invaluable with earlier closure at King’s Cross St. Pancras 2,000 regards to our operations and safety. station. As trains were not stopping there, customers chose to travel On 30 November 2016, Euston station through Euston. was closed at around 18:40 owing to overcrowding. Figure 11 illustrates The result was many more people on 1,500 crowding in different parts of the the platform at Euston, causing severe station just before the closure. The congestion that meant we had to darker areas are the busiest. temporarily close the station to ensure our customers’ safety. Figure 12 shows 1,000 When comparing the amount of people the dramatic increase in people on the in each area during the closure with platform on the day of the disruption, typical conditions, it shows an unusual compared with the usual pattern. number of passengers alighting on the 500 northbound Victoria line and Northern The additional demand led to congestion line (Bank branch) platforms, which in areas of the concourse and on the

0 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 11: Location of crowding levels in Euston before the station closure 30 November 2016 Baseline

Crowding levels Figure 13: Increase in walk times Northern line (northbound) Low Medium High escalators. This was reflected in the walk times between the northbound between the platform and ticket hall Northern line (southbound) platforms and the ticket hall. Figure 13 Concourse (west) shows the median time it took people Walk times increased from to walk from the northbound Northern three minutes at 18:00 to thern line (southbound) Nor line (Bank branch) platform to the more than 10 minutes at ticket hall. The time gradually increased 18:30 because of congestion ) from 18:10 as the station became more oria line (southbound Vict congested, up until it was closed. 18:00 18:15 18:30 Ticket hall Concourse (east) Victoria line (nor thbound)

Northern line (nor thbound)

38 How data can help us improve our services Review of the TfL WiFi pilot 39 Analysing station congestion at this To understand passengers’ route choices, Figure 14: Route options between King's Cross St. Pancras and Waterloo, and level of detail under real circumstances we have relied on customer surveys at the proportion of devices on each one can help us investigate the causes a select number of stations each year. of crowding, understand the effect These ask people about their journey King’s King’s King’s Cross Cross King’s Cross Baker King’s conditions at one station have on other including when and where they started, Cross Street Cross Warren locations, and identify the knock-on which services they used, the lines they Street 6.2% impact of our operational decisions. travelled on, and where they transferred 3.7% Oxford 26.7% Circus 32% It means we can improve the way we between services. The results are then Leicester Green Park Square handle disruptions in the future and modelled to provide a detailed picture of 14.9% minimise the impact on our customers. demand. However, because surveys take time to run and process, we can only use Transport planning certain stations. Waterloo Waterloo Waterloo Waterloo Waterloo Much of the information collected and analysed during our pilot can also This insight helps us to understand King’s King’s Cross Euston King’s Cross King’s help our transport planners improve which stations and lines require capacity Cross Cross Euston 2.4% King’s our network. enhancements. We then run our Cross

transport models to predict how the 1.8% 1.7% Green 2.6% Park Our transport planning teams face a network will operate with the proposed Bank number of challenges, from upgrading improvements in place. We can also the rail network to support a growing predict the impact any works will have on 2% Waterloo London, to looking at what our city will other stations, which informs customer London Waterloo Waterloo Bridge require in the longer term. To do this, communications and operational plans. Waterloo Waterloo they need to understand how customers We use this modelling information to King’s Baker King’s Baker currently travel and how they are likely make changes to timetables so we can Cross King’s Cross King’s Street Cross Street Cross to do so in the future. provide additional capacity and a more

reliable service. 0.2% 1.2% Oxford Using depersonalised ticketing data is a Piccadilly Circus Bank Circus Holborn good way for us to understand demand, As part of our pilot, we tested whether Green Park 1.6% but it only provides an insight into gate- WiFi connection data could provide 1.2% to-gate movements. For some areas, our planners with timely information where stations are served by only one that would help them understand line, this is sufficient. However, much of demand today, and shape planning Waterloo Waterloo Waterloo Waterloo our network is complex – with multiple and modelling in the future. King’s King’s interchange points and routing options, Cross Cross King’s King’s Cross Cross and many paths within stations. In these cases, our ticketing data cannot provide Liverpool Tottenham Street the detailed routing patterns required Court Road Others Oxford Bank when planning our train services and 0.2% Circus 1.4% 0.1% 0.1% station capacity. Westminster London Waterloo Bridge Victoria Waterloo Waterloo Waterloo

40 How data can help us improve our services Review of the TfL WiFi pilot 41 Analysis has shown that the data We can also use the data to get a more provides a wealth of information on accurate understanding of where regular how customers use and interact with crowding within stations occurs. This the network, both under normal can then help identify locations where conditions and during incidents and further investigation needs to take place, disruptions. For example, Figure 14 on particularly if crowding only happens at p41 shows there are at least 18 route specific times or on certain days. By better choices between King’s Cross St. targeting investment, we can provide Pancras and Waterloo. customer improvements more efficiently.

It also highlights the proportion of In addition, WiFi data will enable us to devices connecting to WiFi services on provide aggregate footfall information that route, using a single day's data. The to help generate increased advertising complexity of paths chosen offers far and commercial revenue. more detail about customers’ decisions than we get from our manual surveys. For example, at Westminster station we recently opened a coffee shop The more detailed picture of on the eastbound District and Circle demand that WiFi data offers has the line platform. Before it opened, we potential to support most aspects of carried out surveys and monitoring transport planning, from major station to understand the likely footfall so enhancements to smaller timetable the retailer could decide whether the adjustments. location would be a good investment.

Prioritising investment WiFi data will allow more accurate, round- WiFi data enables us to split stations the-clock observations of potential sites into areas to understand the paths used by measuring the demand and dwell to move between them, and the time time at, and within, stations. This will it takes. We will be able to cut costs by allow us to better understand possible either replacing paper-based surveys locations for future retail opportunities entirely, or by significantly reducing the and help us get the best out of our number we need to carry out. assets so we can raise additional revenue to reinvest in the transport network.

WiFi data can help us identify suitable locations for new retail opportunites, such as this coffee shop at Westminster station

42 How data can help us improve our services Review of the TfL WiFi pilot 43 Conclusion

By collecting and analysing WiFi data Our pilot has been a success both on In view of this, we are now planning we will be able to understand aggregate WiFi data could help a technical and analytical level. We to roll out a network-wide collection customer movements at a given location advertisers target their collected 509 million anonymised process. Given the importance that we and time. This can help companies probing requests in our secure data place on transparency and openness, that advertise at our stations to plan, campaigns system, then created algorithms to we will continue to inform our monitor and measure their campaigns. more effectively, by understand travel patterns. The customers about WiFi connection data showing which station areas pilot showed that the connection collection, the purpose behind it, and As the data will also enable us to have the greatest footfall data generated as a by-product of what it will be used for. We will also identify the areas in our stations with our WiFi services can offer substantial follow the ICO’s guidelines and rules, the greatest footfall, we can understand benefits for us and our customers. and explain how customers can opt out which sites are most valuable to by switching off WiFi on their devices. companies wanting to advertise. Using Analytic tools and services built from our digital advertising screens, where this connection data can improve the We now look forward to taking the next we have dynamic control over content, way we run and plan our network, steps to introduce this new technology we can target advertising in the right and can provide our customers with to benefit our customers. place at the right time to maximise much more detailed information additional income. We will be able to about their journeys. frequently, and rapidly, measure how many customers could have viewed the adverts to enable companies to review their campaigns. And as WiFi can be used to split a station into many distinct areas, we can provide a detailed picture.

44 How data can help us improve our services Review of the TfL WiFi pilot 45 © Transport for London Windsor House 42 – 50 Victoria Street London SW1H 0TL August 2017 tfl.gov.uk PUB17_031