Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
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.
500 million
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
to look at journey patterns. However, Ahead of the launch of the pilot we
while this data is very useful for
depersonalised WiFi connection requests Victoria carried out a press and awareness
understanding gate-to-gate patterns, Northbound
Northbound Trains arriving
Trains arriving 90%
90% full
full campaign. We also pledged to share the
it does not tell us which routes Southbound
Southbound Trains arriving
Trains arriving 75%
75% full
full 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 Northbound Trains arriving 90% full
The pilot was successfully delivered,
Northbound
our ticketing data with manual customer this changes in response to events Southbound
Trains arriving 50% full
Trains arriving
arriving 75%
using in-house expertise, for less than
Southbound Trains 25% full
full
surveys. This is time intensive, expensive and network alterations £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.
WiFi connection data could help us
From our data analysis, we have been data to understand travel patterns
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.
With 45 per cent of the Tube network Figure 1: Map of the pilot area
Harrow-
on-the-Hill
Preston
Road
Hendon Central Highgate Seven Tottenham Blackhorse
Kingsbury Hale Road
in tunnels and stations as deep as 58.5 Rayners Lane
West Northwick
Brent Cross
Gospel
Archway
Manor House
Sisters
Walthamstow
Central
Redbridge
Park
metres, there are many areas where Harrow
South Kenton Neasden
Golders Green
Hampstead
Heath
Oak Upper Holloway Stamford
Hill Walthamstow
Wanstead
Leytonstone
Gants
Hill
Wembley Park Hampstead Leyton
cellular networks do not provide South Harrow North Wembley
Dollis Hill
Tufnell Park Arsenal
Finsbury Stoke
Queen’s Road
Midland Road
Wembley Central Park Newington
connectivity. In partnership with Virgin Stonebridge Park
Willesden Green Finchley Road
& Frognal Belsize Park
Kentish
Town West Kentish Town
Holloway Road
Rectory
St. James Street Leytonstone
High Road
Wanstead
Park
Forest
Road
Media, we started installing WiFi in Sudbury Hill
Harlesden
Kensal Brondesbury
Kilburn
West Chalk Farm Camden
Caledonian Road
Highbury &
Islington
Dalston
Kingsland
Clapton Leyton Gate
Stratford
stations in 2012, and it is now available Willesden Junction Rise Park Hampstead
Camden Town
Road
Caledonian
Hackney
Downs Hackney
Central
International Maryland
Street
up-to-the-minute travel information. Perivale
Maida Vale
Warwick
Avenue
Street
Angel
Hoxton Bethnal Green Pudding
Mill Lane
Abbey
Road Upton
Park
Old Street Bethnal Green Mile End
Euston Bow
Edgware Warren Street Square Road Plaistow
Royal Oak Road
Farringdon Shoreditch
Connectivity data is generated as a Hanger Lane
Westbourne Park
Regent’s Park Russell
Square
High Street Stepney Green Bromley-
by-Bow
West Ham
Bow
by-product of providing this service. Park Royal Ladbroke Grove Bayswater
Bond Oxford
Goodge
Street
Barbican
Aldgate
Church
East Whitechapel
We already use a range of data, such as North Ealing
Latimer Road
Marble Arch
Street Circus
Tottenham Chancery Lane
Moorgate Liverpool
Street
Devons Road
Star Lane
Holborn Langdon Park
aggregated and depersonalised Oyster East
Acton
White
City
Shepherd’s
Bush
Notting
Hill Gate
Court Road
Bank
Aldgate
Covent Garden All Saints Canning
and contactless payment transactions, Ealing Broadway
West
Acton
North
Acton
Holland
Park
Queensway
Lancaster
Gate
Green Park Leicester Square
St. Paul’s
Limehouse Poplar
Town
Wood Lane
to understand how customers travel Acton Central Shepherd’s
Hyde Park Corner
Piccadilly
Circus
Cannon Street Tower Shadwell Westferry Blackwall East
High Street Kensington Monument India
across London. While this tells us Ealing Common
South Acton
Bush Market Kensington
(Olympia) Knightsbridge Charing
Mansion House
Hill
Fenchurch Street Tower
Gateway Wapping West India
Goldhawk Road
where customers enter and exit the South
Ealing Acton Town Hammersmith
Barons
Court
Gloucester
Road St. James’s
Cross
Blackfriars River Thames
Quay
Rotherhithe
Tube network, it does not reveal Victoria Park Temple Canary Wharf
Emirates
anything about their route choices, Northfields Chiswick
Park
Turnham Stamford Ravenscourt
Green Brook Park
West
Kensington
Earl’s
Court
South
Kensington
Sloane
Square
Westminster Embankment London
Bridge
Bermondsey Canada
Water Heron Quays
North
Greenwich Greenwich
Peninsula
and around stations, and how they Fulham Broadway Lambeth North Island Gardens
Richmond Parsons Green
responded to network changes. Putney Bridge
Imperial Wharf
Cutty Sark for
Maritime Greenwich
Queens Road
East Putney Peckham New Greenwich
Owing to the layout of the Tube Southfields
Vauxhall
Elephant & Castle Cross Gate
New Cross Deptford Bridge
Clapham North
us plan and operate the network, Clapham Common
Sydenham
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
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.
Scenario 1: Road beacons – for real-time traffic flow and congestion information Scenario 3: WiFi connection data on the Tube – for better customer information Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
96601 96601
Victoria
Northbound Trains arriving 90% full
Southbound Trains arriving 75% full
Jubilee
Northbound Trains arriving 50% full
Southbound Trains arriving 25% full
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
1. Next train Caution
Sliding doors
Caution
Sliding doors
1 min Caution
Sliding doors
Victoria
2. Next train 3 mins Northbound Trains arriving 90% full
Jubilee
Northbound Trains arriving 50% full
Southbound Trains arriving 25% full
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
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 and Earl’s Court. We noticed that you
station on Pall Mall A2 roadworks have didn’t tap out at
is currently empty Delay on number 8 Wednesday, 14:39
Morden yesterday
been completed early
but a bike delivery is bus route between morning, so we have
– all lanes will be
due at 11am. Sorry for St. Pauls and Holborn. auto-completed your
reopened from 10am
38 any inconvenience. journey to avoid you
being charged the
Yours sincerely,
maximum fare.
pilot launches in
Shashi Verma, chief technology
officer at TfL, said: ‘This short trial Opportunity
Tube stations.
will help us understand whether WiFi
connection data could help us plan
and operate our transport network
to opt out
STATION
The ride, which takes place on
Sunday July 30, offers cyclists the
chance to experience some of London’s
most historic streets, free from cars,
benefits of doing so. For more articles and to keep up to date with TfL announcements, visit tfl.gov.uk/news
The week before launching the pilot, call 020 3176 2999, or visit
www.londontravelwatch.org.uk
we issued a press release1 that set out * Service and network charges apply.
See tfl.gov.uk/terms for details.
the benefits and scope of the project. Newspapers left on the Tube can jam doors and cause delays to your journey.
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.
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
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
509
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
contains an identifier specific to that
device, known as a Media Access Control
did not subsequently do so. Once the
device had connected, we also gathered million
(MAC) address. If a WiFi network is these requests and authentications. probing requests
found that is known to the device, it will No web browsing data or data from were collected at Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
54
automatically connect. If the device finds website cookies was gathered. 96601 96601
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
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
25,000,000
King’s Cross St. Pancras generated the devices, we either collected single
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 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
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 pilot25
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.
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
AB
Edit journey Add favourites Edit preferences ˅
For example, someone planning to Many customers want to understand available through our travel tools. It 2 mins the Victoria line between Euston
Northern line to Embankment Underground Station
6 mins
travel between Camden Town and how busy the train(s) and interchange highlights the average
Departs Arrives demand across and Green Park.
Victoria line to Victoria Underground Station
1 min
St. James’s Park is presented with station(s) could be so they know if they each section13:07 13:22two journeys
of these 15 mins District line or Circle line to St. James’s Park Underground Station
two options when using TfL Journey are able to board the first train. Data during the morning peak period
View Details
i This crowding can be avoided by taking
This journey has additional information
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 crowded 10 mins
direct to Embankment. Some customers
Northern line to Embankment Underground Station
These options, however, do not help level of detail. train in the first option, this is quicker 2 mins may always prefer the quicker of the
District line to St. James’s Park Underground Station
Departs Arrives
them make their decision based on although significantly
13:10 13:27more crowded, 17 mins i
two options, but by providing these
This journey has additional information
factors other than estimated time Figure 7 shows how we could use WiFi especially on the branch between
View Details
details we can help them make
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
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
Departs Arrives Departs Arrives Goodge
6 mins Victoria line to Victoria
1 min 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 Oxford Circus
1 mins District line or Circle line to
View Details
i This journey has additional information
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
2 mins Departs Arrives 2 mins District line to St. James’s
District line to St. James’s Park Underground Station Cross
Departs Arrives 13:10 13:27 17 mins Park Underground Station Victoria
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
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
for individual trains. at station and line level, allowing 7:50-7:55
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
100+ passengers
people to make decisions about the
While train-level demand remains route they take. We currently provide
relatively consistent across peak periods, information on the busiest times, but 7:55-8:00 Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
300+ passengers
there are some fluctuations in crowding WiFi connection data would allow us
levels between neighbouring trains. to provide greater detail.
This is partly driven by the number of 8:00-8:05 Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
to board, and the number alighting paths customers take within stations, and 8:05-8:10 Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
gives us a detailed picture of the busiest 8:10-8:15 Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
700+ passengers
If we can provide information for areas at different times of the day.
crowding levels on specific trains,
customers may choose to wait for the Figure 9 on p34 shows how customers 8:15-8:20
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
service for everyone by smoothing Victoria line southbound platform at 8:20-8:25 Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
the busiest times on the morning of more complex paths. Those who use the 8:30-8:35
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
08:20 and 08:25, three trains departed and those who use the concourse take 8:35-8:40
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
trains departing in the preceding and want to understand transfer times and 8:40-8:45
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
Sliding doors
Caution
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
o u thbound)
i n e ( s
Nor thern l
o uthbo und)
l i n e ( s Customers choose different routes through
Victo ria stations, depending on their needs
When choosing a route through a station, How does disruption affect my journey?
Victoria customers have different requirements Another potential benefit is the
li ne (nort and preferences. Detailed information responsiveness of the data to real-time
3-5 hbound) 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
Norther
1-3 n line (n
orthbou
nd)
travelling with buggies or luggage, would
benefit from knowing the least crowded
disruption. We then make assumptions
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
customer reactions. Embankment towards Monument.
08:00-09:00 Waterloo & City line suspended, which led to
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
Leicester Square
of the most crowded sections on the many more people changing here
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
Waterloo Around 4,000 decided to use the Jubilee
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 as it enables a more informed response
Suspended Waterloo & City line
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
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
Waterloo
on the Jubilee line to Westminster. trains to ensure the quickest and most
Westminster All three of these responses meant a comfortable alternative journeys.
Monument/Bank
Waterloo
Westminster
London Bank
Bridge
Waterloo
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
Norther
n line (n
orthbou
nd)
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
Cross
King’s
Cross King’s
King’s
Cross Baker King’s
conditions at one station have on other including when and where they started, Cross Street Cross
locations, and identify the knock-on which services they used, the lines they Warren
Street
6.2% 3.7%
impact of our operational decisions. travelled on, and where they transferred Oxford 26.7%
It means we can improve the way we between services. The results are then Circus 32%
Leicester
Green Park
handle disruptions in the future and modelled to provide a detailed picture of Square
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
to-gate movements. For some areas, our planners with timely information
where stations are served by only one that would help them understand
Waterloo Waterloo Waterloo Waterloo
line, this is sufficient. However, much of demand today, and shape planning
our network is complex – with multiple and modelling in the future.
King’s King’s
interchange points and routing options, Cross Cross King’s
Cross
King’s
Cross
and many paths within stations. In these
cases, our ticketing data cannot provide Liverpool
the detailed routing patterns required Tottenham
Court Road
Street
Others
Oxford Bank
when planning our train services and 0.2% Circus 1.4%
0.1%
station capacity. Westminster
0.1%
London
Waterloo Bridge
Victoria
Waterloo Waterloo Waterloo
40 How data can help us improve our services Review of the TfL WiFi pilot41
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.
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
that advertise at our stations to plan,
monitor and measure their campaigns.
campaigns probing requests in our secure data
system, then created algorithms to
understand travel patterns. The
place on transparency and openness,
we will continue to inform our
customers about WiFi connection data
more effectively, by
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 pilot45
© Transport for London
Windsor House
42 – 50 Victoria Street
London SW1H 0TL
August 2017
tfl.gov.uk PUB17_031