Crowd Sourcing

This measure was provided by INSTITUTE FOR TRANSPORT STUDIES (ITS) in 2014 under the CH4LLENGE project, financed by the European Commission.

---

Crowd sourcing can work in two different ways. The first is through the passive/semi-passive collection of information via Web 2.0 enabled devices such as smart phones which may be being carried by travellers on public transport services or in cars on the road network. This can be enriched by asking the crowd to provide supplementary information such as what mode they are travelling on and if public transport what the service number is.

Passive and semi-passive crowd sourcing is similar to data collection undertaken by public transport companies via Automatic Vehicle Location (AVL) systems.  Such AVL systems are used to update real time travel information at public transport stops/stations and operator websites.  The cost of AVL systems can be expensive and crowd sourcing offers a more cost effective way to collect similar data via third parties, namely transport users.

The second method of collecting data is pro-active and invites the crowd to generate opinions and perspectives upon the service they are experiencing.  For example, users may generate data on the level of crowding on a bus/train or of congestion.  Such information can then be disseminated via the crowd sourcing application and combined with the passive information being provided, so that users not only know whether a bus/train is operating to schedule but also whether they are likely to obtain a seat when they board.  This may influence their decision on whether to board that particular bus/train or wait for a less congested vehicle.  This in effect is the best of both worlds and offers something that current AVL systems cannot.

Crowd sourcing offers the potential for strong efficiency gains for both public transport and road users by enabling them to make smarter choices – by reducing waiting time at public transport stops/stations, avoiding congested routes and overcrowded public transport services.  From a financial viewpoint crowd sourcing can offer substantial cost savings on similar systems (e.g. AVL) especially from the infrastructure side as data is provided by users of Web 2.0 devices as opposed to a purpose built stationary sensor. 

There are concerns about the use of crowd sourcing, centring on the quality of the data it provides (this is reliant on the number of users, the quality of the data they provide and its interpretation) and the potential exclusion of certain groups.  The latter may include those on low incomes who cannot afford a Web 2.0 device and those who are not technology-aware (e.g. the elderly).  Such concerns need to be addressed jointly by application developers and transport operators/authorities.

Introduction

Various definitions exist for the term crowd sourcing.  Several studies employ different terms with the following being one of the most comprehensive.

“Crowd sourcing is a type of participative online activity in which an individual, an institution, a non-profit organisation, or company proposes to a group of individuals of varying knowledge, heterogeneity and number, via a flexible open call, the voluntary undertaking of a task” (Arolas & Guevara, 2012)

In reference to travel information the term actually refers to ‘ubiquitous’ crowd sourcing (Mashadi & Capra, 2011) in that the crowd sourcing process is dynamic and reliant on Web 2.0 users (i.e. smart phone users) who are responsible for producing real time updates on travel information that are then consumed either by fellow Web 2.0 users or by users accessing pre-trip information from a static internet device

Terminology

Crowd sourcing is a generic term applied to an act which involves engaging with a ‘crowd’ (e.g. public transport users or motorists) to ‘source’ information (e.g. reliability of public transport or congestion levels on roads).  Crowd sourcing falls into two categories, ‘Web-based’ and ‘Ubiquitous’.  The former refers to static crowd sourcing whereby the crowd is either engaged to complete a task, (e.g. Amazon Mechanical Turk1) or to report problems, such as potholes and graffiti to local government (e.g. www.fixmystreet.com).  The latter refers to dynamic crowd sourcing, whereby users provide information in real time which is gathered, analysed and then fed back to the same users and a wider audience.

Description

Ubiquitous crowd sourcing (henceforth referred to as crowd sourcing) can work in two different ways, with both ways often combined.  The first is through the passive collection of information, which may involve the generation of information by Web 2.0 enabled devices such as smart phones.  If a person enables their device to send out GPS information which can be picked up by a crowd sourcing application, then that data provides a log of the progress of that person’s journey either by car or by public transport.  Purely passive data collection can create a lot of ‘noise’ since it might not be clear what transport mode the crowd are travelling on.  To counter this some crowd sourcing systems, such as TIRAMISU (http://www.tiramisutransit.com/), ask the crowd to provide supplementary information such as what mode they were travelling on and if public transport what the service number was.  This makes the data much more accurate compared with purely passive data collection. 

Passive and semi-passive crowd sourcing is similar to data collection undertaken by public transport companies via Automatic Vehicle Location (AVL) systems.  Such AVL systems are used to update real time travel information at public transport stops/stations and operator websites. The cost of AVL systems can be expensive and passive crowd sourcing offers a more cost effective way to collect similar data via third parties, namely transport users.

The second method of collecting data is pro-active and invites the crowd to generate opinions and perspectives upon the service they are experiencing.  For example, users may generate data on the level of crowding on a bus or a train.  Such information can then be disseminated via the crowd sourcing application and combined with the passive information being provided, so that users not only know whether a bus/train is operating to schedule but also whether they are likely to obtain a seat when they board.  This may influence their decision on whether to board that particular bus/train or wait for a less congested vehicle.  This in effect is the best of both worlds and offers something that current AVL systems cannot.

A good example of this combined crowd sourcing is the TIRAMISU smartphone app which was launched by researchers at Carnegie Mellon University (EMARQ Network, 2011) and was initially aimed at crowd sourcing public transport services within the Pittsburgh (U.S) area. Since its launch the service has focused on bus services and been expanded to provide coverage to bus travellers in Syracuse and New York City.  When opened the app shows a list of the nearest bus stops and displays the arrival times of buses based upon data obtained from real time crowd sourcing (or in its absence the Port Authority schedule).  The TIRAMISU app relies on its users ‘logging in’ and highlighting what bus service they are on. Users are also encouraged to indicate the levels of crowding on buses by choosing one of four levels (many seats, few seats, no seats or full).  This is seen as particularly helpful for people with mobility issues but is also beneficial for passengers in general.

Other examples from a public transport background include www.hopstoplive.com and www.moovittapp.com which provides similar real time information but on a much bigger (international) scale and which combine the service with more static information such as door to door journey planning, maps and real time updates such as arrival notification and/or get off alerts for your stop.  Applications also exist for car users with www.waze.com being the largest. Recently acquired by Google for $1.1billion (Rao, 2013) with a view to integrating into Google Maps, Waze provides live traffic data for drivers as well as the opportunity for users to actively share information on accidents, speed cameras and other road incidents.

Why introduce crowd sourcing?

There are two main rationales for using crowd sourcing to produce travel information that relate to the key strengths of crowd sourcing.

1. Cost savings – they offers substantial cost savings since the information is being provided by the crowd rather than an AVL system.  This offers the opportunity for greater coverage of real time travel information (for both road and public transport networks) via the roll out of crowd sourcing apps either by government or transport operators or by other entities, e.g. MOOVIT.
2. Richer content – the pro-active nature of the data collection and the fact it is being provided by actual users in real time means that the data provided has much more depth (e.g. levels of crowding) and in theory more accuracy making it much more valuable to current and potential users.

Paradoxically these strengths may also be the source of weaknesses and barriers to their continued development, in that:

1. The size of the crowd – is crucial to the level of coverage that crowd sourcing apps can provide.  In order to provide full coverage there must be at least one enabled user on every public transport service at all times.  Currently this is not the case, with inevitable gaps in the off-peak and inter-peaks.  Over time it is hoped that additional users will come online to fill any gaps but this is a classic ‘chicken and egg’ situation in that users will only join if they perceive the information being provided is comprehensive and that will only be the case when more users join.
2. Quality of the content – may be subject to the vagaries of subjective opinions and/or incorrect information (e.g. just how crowded is the bus, just how bad is the congestion, which bus service am I on).  If no quality controls are in place then incorrect data can be misleading to other users and lead them to lose trust in the crowd sourcing app.
3. Architecturally - it is important that systems can speak to each other and that information can be shared.  Currently a number of apps are available which are competing against each other rather than pooling data to ensure a better coverage for all users.
4. Equity issues – may arise if certain groups cannot access crowd sourcing data, especially if governmental/transport operators decide not to provide live transport information at stops/stations and instead make crowd sourcing apps the de facto source of live travel information.

Demand impacts

The demand impacts of crowd sourcing travel information are likely to be minimal in terms of changing the overall demand for travel.  People are still likely to make the same number of trips to the same destinations but will make smarter choices in terms of adjusting their time of travel to avoid congested modes/services and overcrowded services.  For large cities, such as London, where choice between public transport modes is possible, it is likely that crowd sourcing will lead to greater levels of switching between public transport modes to avoid delays and overcrowding. Any reduction in road traffic will be minimal at best.  Longer term car users may switch to public transport if they feel more informed about the reliability of public transport operations.

Responses and situations
Response	Reduction in road traffic	Expected in situations
		Both car and PT travellers may consider changing their departure time to avoid congestion but this will not reduce road traffic.
		Both car and PT travellers may consider changing route if they have better real time information (rich in content) across all modes & more networks but this will not reduce road traffic.
		If very severe congestion is flagged up then it a change of destination may occur (both car and PT users). However, the frequency of such events means their impact on road traffic will be minimal.
		-
		The availability of improved real time information (rich in content) across all modes and wider networks will ensure more informed travellers.  This may persuade car users (in the long term) to make some trips by PT.
		-
		-

	= Weakest possible response		= Strongest possible positive response
	= Weakest possible negative response		= Strongest possible negative response
	= No response

 

Short and long run demand responses

There is the possibility of stronger reductions in road traffic if the provision of high quality, reliable and easy to access PT travel information continues and provides reassurance to car users that they can use PT more effectively.  Any such reductions however are likely to be small.

Demand responses
Response	1st year	2-4 years	5 years	10+ years

	= Weakest possible response		= Strongest possible positive response
	= Weakest possible negative response		= Strongest possible negative response
	= No response

Supply impacts

It is unlikely that there will any impacts on supply as a direct result of this instrument.

Financing requirements

It is acknowledged that using crowd sourcing reduces the costs of providing the infrastructure of an AVL system, which for a mid-sized city maybe as much as $70 million (Zimmerman et al, 2011).  Despite this there are a number of other costs that need to be highlighted and taken in account (Di Maio, 2009), these are:

1. Initial software engineering costs - associated with developing the software application and testing/piloting it.
2. Ongoing software engineering costs - associated with maintaining the software application and ensuring its smooth operation.
3. Promotion/marketing costs - associated with launching the application and encouraging take up amongst users.  Whilst crowding sourcing apps may rely on user recommendations to increase uptake there need to be ongoing advertising campaigns to ensure continued growth through new channels.
4. Operational costs - will be similar to any business model but will be slanted towards an IT perspective, e.g. cost of monitoring and checking data readings/ratings.

Attempting to place values on these costs is difficult and will clearly vary from application to application depending upon the scale of the operations, the aims of the company developing the app and the exact services they offer.  Lauvray (2011) notes that the development cost of a business app that includes real time data can amount to $150,000 plus.  Moovit recently announced it had secured funding of $28 million (Rao, 2013) after previously receiving $3.5 million to help fund its start up in 2012.  With 3 million users in over 100 cities this gives an idea of the scale of the operations and the costs associated with it.  It is unlikely, that smaller scaled applications such as TIRAMISU would have costs approaching that of Moovit and there are opportunities to use the ‘crowd’ to take on some of the operational roles (such as managing twitter accounts and support emails, Mott - 2013). 

Expected impact on key policy objectives

Contribution to objectives
Objective	Scale of contribution	Comment
		Both PT and car travellers will gain large efficiency gains by using live information (with rich content) to make smarter travel choices in times of when to travel and by which mode.  This should reduce both wait and journey times.
		No direct impact expected.
		No direct impact expected.
		Possible social exclusion issues if certain segments of the population are not able to access the information provided either via the web or via Web 2.0 technologies.
		No direct impact expected.
		No direct impact expected.
		Whilst it is difficult to quantify costs the recent trends would suggest that the costs of providing crowd sourcing travel information is being borne by the private sector and so reducing the burden on government and transport operators. From an infrastructure perspective the technology employed offers substantial cost reductions from comparable technologies such as AVL.

	= Weakest possible positive contribution		= Strongest possible positive contribution
	= Weakest possible negative contribution		= Strongest possible negative contribution
	= No contribution

Expected impact on problems

Contribution to alleviation of key problems
Problem	Scale of contribution	Comment
Congestion		Congestion levels should be reduced by enabling car users to make smarter travel choices in terms of where they travel and when.
Community impacts		No direct impacts expected.
Environmental damage		A possible small impact on environmental damage if car users make smarter travel choices which avoid stationary engine running.
Poor accessibility		No direct impacts expected.
Social and geographical disadvantage		No direct impacts expected.
Accidents		No direct impacts expected.
Economic growth		No direct impacts expected.

	= Weakest possible positive contribution		= Strongest possible positive contribution
	= Weakest possible negative contribution		= Strongest possible negative contribution
	= No contribution

Expected winners and losers

Winners and losers
Group	Winners/Losers	Comment
Large scale freight and commercial traffic		May benefit from reduced congestion levels if car users make smarter travel choices.  May also be some benefits for themselves if they are able to avoid congestion by altering their time of travel &/or congested routes.
Small businesses		May benefit from reduced congestion levels if car users make smarter travel choices.  May also be some benefits for themselves if they are able to avoid congestion by altering their time of travel &/or congested routes.
High income car-users		All car users should benefit but car users might benefit slightly more than other users if they have greater access to Web 2.0 devices such as smart phones.
Low income car-users with poor access to public transport		Dependent upon whether low income car-users have the same level of access to Web 2.0 devices.
All existing public transport users		Likely that PT users will be winners.  The extent of the gains will be decided by the choices of different public transport available to them and also the access to Web 2.0 devices.
People living adjacent to the area targeted		No direct impacts expected.
Cyclists including children		No direct impacts expected.
People at higher risk of health problems exacerbated by poor air quality		No direct impact expected, unless emissions reduced.
People making high value, important journeys		Given that these are likely to be business trips then there is a likelihood of considerable gains by avoiding congestion either as a car or PT user.
The average car user		Will enjoy some benefit from reduced congestion from route switching or trip retiming.

	= Weakest possible benefit		= Strongest possible positive benefit
	= Weakest possible negative benefit		= Strongest possible negative benefit
	= Neither wins nor loses

Barriers to implementation

Scale of barriers
Barrier	Scale	Comment
Legal		No obvious barrier.
Finance		Current developments appear to be driven by the private sector (Moovit) or by smaller academic/community based start ups (TIRAMISU) so no barriers in terms of public financing constraints.  The use of Web 2.0 devices & their users to act as information gatherers has reduced infrastructure costs substantially.  Similarly, the active involvement of the crowd to help operate some aspects of the service (twitter accounts) also reduces the operating costs.
Governance		No obvious barrier.
Political acceptability		No obvious barrier.
Public and stakeholder acceptability		Some possible concern from the public about the quality of the information being provided and the availability of apps across all platforms.
Feasibility		Some issues related to data quality and coverage.  The latter will be resolved with the growth of and acceptance of the technologies involved.  Also some concern as to whether apps will be developed for all potential platforms.

= Minimal barrier

= Most significant barrier

MOOVIT (Worldwide) & OneBusAway (US)

Moovit App

OneBusAway Screen Shot

Context

The two case studies are now discussed jointly:

Moovit is a private sector company founded in 2011 in Israel (Rao, 2013). It offers a real time public transport information service to users via Web 2.0 devices (smart phones).  A range of public transport services are covered including buses, trains, trams, metro/underground systems, trolleybus and river ferries, tailored to the PT services surrounding the user.  Moovit combines PT schedules with real time data collected from PT users using crowd sourcing techniques, e.g. collecting real time data passively from smart phones (Purdy, 2012).  It is the latter that provides it with a USP, enabling it to provide real time data to its users.  Moovit also allows the ‘crowd’ to report additional data which adds richer content for PT users, such as levels of crowding, reasons for delays and the likely severity etc. Additional features include a journey planner and maps of stops, stations and the surrounding areas. 

Where possible Moovit works with established transport providers or transport authorities (http://www.moovitapp.com/all-uk-england/) and has grown rapidly boasting a presence in 350 cities (Dickey, 2014) and 6.5 million users (Forbes, 2014). Architecturally, Moovit is a closed system, which means it retains full control of the system and can choose which cities it operates in (Barbeau et al., 2013).  In contrast, the OneBusAway application is an open source software, meaning that it is available to ‘anyone and everyone’.  In this way, it is not reliant on the whims of one software company.  If one regional operator decides to discontinue providing the information service another operator can start it up. This is an important advantage for OneBusAway.  Another is that the system is available on four different mobile app platforms (Android, Windows Phone, Windows 8 and iPhone), unlike Moovit which is available as Android and iPhone only.

OneBusAway was conceived by students at the University of Washington, Seattle in 2008.  It initially spread virally to around 100,000 users as an unofficial real time system before Sound Transit, King County Metro and Pierce Transit (all in and around Seattle, U.S.) provided financial support between 2011 and 2013 (Barbeau et al., 2013).  The key drivers behind the OneBusAway system is not just the development of its open source software, but also the emergence of open transport data, in particular data standards, e.g. GTFS – originally created for Google Transit trip planner. With over 500 transport agencies offering open access to their PT schedules this has allowed the creation, by third parties, of travel applications (City-Go-Round, 2014).  One of the main rationales for the development of the OneBusAway application was to provide a body of research to test and assess the impacts of real time information on user behaviour.  That is still the case today and it is a very different product from the more commercially driven Moovit.  As an application it is therefore more limited than Moovit with its principal focus still to provide real time information, whereas Moovit is developing into a ‘one stop information shop’ with the provision of complementary services like a journey planner.

Impacts on demand

Whilst there is literature and evidence on the demand impact from real time information systems per se (as reported elsewhere in KonSULT), there is little evidence on the impact of demand from specific crowd sourcing applications like Moovit and OneBusAway.  What evidence there is centres on research carried out by Gooze et al. 2012.  They examined the impact of the OneBusAway system in the Seattle region with regards to ridership and other benefits.  Their methodology compared the results of a survey carried out in 2009 (1 year on from the introduction of OneBusAway)  with a repeat survey carried out in 2012, in the hope of identifying statistical differences between the two that could be attributed to the OneBusAway system.

In terms of demand two distinct patterns appear to have emerged.  The research reports a significant and positive impact in trip making as a direct result of the OneBusAway system with around 12% more users in 2012 vis a vis 2009 making at least one more trip than they used to.  This is put down to the rider experience providing positive reinforcements about the reliability of bus, which in turn encourages existing users to travel more by bus (either making new trips or shifting mode). 

Countering this positive impact on demand is a negative one associated with errors in the real time data information conveyed to users.  The introduction of the OneBusAway system appears to have driven up expectations amongst users with regard to reliability and in particular the performance of the real time system.  With 77% of users experiencing a real time prediction error there is danger that the positive trip generation associated with the introduction of real time crowed sourced information may be eroded.  Both transport agencies and application developers need to tackle the issue jointly. 

Whilst there is no direct evidence on mode shifting, the re-timing of journeys or the re-routing of journeys, all three are expected to occur, especially the latter two.

Impacts on supply

There was no evidence on changes in supply but evidence from AVL systems as reported in KonSULT would suggest that real time information systems can lead to significant patronage increases which may prompt bus operators to increase frequencies.  This is however a case by case basis and it should be borne in mind that as bus frequencies cross a certain threshold (e.g. every 10 minutes) this itself starts to negate the benefits of having real time data.

Contribution to objectives

Contribution to objectives
Objectives	Scale of contribution	Comment
		There would appear to be efficiency gains for both users and the operators. Users reduce the amount of time they waste waiting at the bus stop, whilst operators can have more confidence in that they can operate a tighter service pattern with fewer vehicles.  The former is eroded by errors in the real time system which need to be controlled.
		No ‘hard‘ evidence to suggest that real time crowd sourced systems are taking car drivers from the street but it remains a possibility as public transport begins to be perceived as a more reliable mode.
		As above no ‘hard’ evidence to suggest significant mode shifting away from car, but it remains a possibility.
		Some concern expressed in the literature that potential equity issues exist and needs to be considered in the design of any system.  Everyone should have access to this information regardless of location, income or platform.  This may not be the case for lower income users who do not have access to Web 2.0 devices.
		Evidence to suggest that users of PT feel safer (32% of users), knowing when a bus/train is going to arrive.  This relates to a reduction in time spent waiting at a stop/station - locations which people might feel more vulnerable.  This might be classed as improving security rather than safety.
		No evidence on economic growth was found.
		Evidence that this application could bring major cost savings since a normal AVL system for a mid-sized city could cost $70 million and the commercial development of a business application is around $150,000.

	= Weakest possible positive contribution		= Strongest possible positive contribution
	= Weakest possible negative contribution		= Strongest possible negative contribution
	= No contribution

Contribution to objectives and alleviation of problems

Contribution to objectives
Objective	Scale of contribution	Comment
		An expectation that both PT users and car travellers will enjoy large efficiency gains from being able to make smarter travel choices which reduce their wait time at the PT stop/station and whilst travelling.  The evidence supports the former, and also indicates efficiencies for operators.
		No real expectations of improving the liveability of streets and no evidence either.
		No expectations of improving the protection of the environment and no evidence either.
		Some concerns that certain segments of society (those on low income or who are not technologically aware) may be excluded.  No hard evidence to suggest this is the case but circumstantial evidence to suggest it is.
		No prior expectations that there would be a contribution to safety. The evidence supports this with regard to accidents but not when it comes to personal security - with evidence that there is some benefit in this area.
		No expectations of changes in this area and no evidence to suggest otherwise.
		Expectations that some major costs savings may arise - particularly in the area of infrastructure.  Evidence shows this is the case.

	= Weakest possible positive contribution		= Strongest possible positive contribution
	= Weakest possible negative contribution		= Strongest possible negative contribution
	= No contribution

 

Contribution to problems
Problem	Scale of contribution	Comment
Congestion		No expectation or evidence to suggest that PT user applications impact upon congestion.  Some expectation that road user applications will reduce congestion.
Community impacts		No expectation of impacts and no evidence.
Environmental damage		No expectation of impacts and no evidence.
Poor accessibility		-
Social or geographic disadvantage		-
Accidents		No expectation of impacts and no evidence.
Economic growth		No expectation of impacts and no evidence.

	= Weakest possible positive contribution		= Strongest possible positive contribution
	= Weakest possible negative contribution		= Strongest possible negative contribution
	= No contribution

Appropriate contexts

Appropriate area-types
Area type	Suitability
City centre
Dense inner suburb
Medium density outer suburb
Less dense outer suburb
District centre
Corridor
Small town
Tourist town

= Least suitable area type

= Most suitable area type

Adverse side effects

There are concerns that certain segments of society, namely those on low incomes or who are not technologically aware (such as the elderly) may not be able to take advantage of the new applications. Care must be taken by transport operators and authorities that such groups are not excluded (e.g. ensuring that the information produced is available via the internet and not just Web 2.0 devices) and can share in the benefits.

Another set of concerns are raised on the increasing sensitivities of users to errors in real time information.  Quality thresholds need to be continually improved if the benefits accruing from crowd sourced information applications are not to be eroded at the edges.

Barbeau, S., Borning, A. and Watkins, K. (2013) OneBusAway Multi-Region – Rapidly Expanding Mobile Transit Apps to New Cities. http://www.slideshare.net/sjbarbeau/onebusaway-multiregion-rapidly-expanding-mobile-transit-apps-to-new-cities-nctr-advisory-board2

Bertoni, S. (2014) http://www.forbes.com/sites/stevenbertoni/2014/06/09/lyft-partners-with-moovit-as-car-sharing-battle-continues/

City-Go-Round (2014) http://www.citygoround.org/agencies/

Dickey, M.R. (2014) http://www.businessinsider.com/lyft-partners-with-moovit-2014-6

DiMaio, A. (2009) http://blogs.gartner.com/andrea_dimaio/2009/06/04/whats-the-real-cost-of-crowdsourcing/

EMBARQ Network (2011). http://www.crowdsourcing.org/document/crowdsourcing-realtime-transit-updates/6116/3

Estellés-Arolas, E. and González-Ladrón-de-Guevara, F. (2012) Towards an integrated crowdsourcing definition.  Journal of Information Science 2012 (38) 189-200.

Gooze, A, Watkins, K., Borning, A. (2013). “Benefits of Real-Time Information and the Impacts of Data Accuracy on the Rider Experience.” Proceedings of the Transportation Research Board 92nd Annual Meeting.

Lauvray, L. (2011) "iPad App Development Cost - A Breakdown." Propelics Blog, Vol.

16 Accessed at: http://www.propelics.com/ipad-app-development-cost-a-breakdown/

Mashadi, A. and Capra, L. (2011) Quality Control for Real-time Ubiquitous Crowdsourcing. Proceedings of the 2nd International workshop on Ubiquitous Crowdsourcing.

Mott, N. (2013) http://pando.com/2013/09/18/moovit-expands-by-relying-on-the-crowd/

Purdy, K. (2012). http://www.fastcompany.com/3004264/moovit-wants-fix-your-nightmare-commute

Rao, L. (2013). http://techcrunch.com/2013/12/18/moovit-raises-28m-from-sequoia-and-others-to-be-the-waze-for-crowdsourced-public-transit-data/

The Verge (2013). http://www.theverge.com/2013/4/8/4195266/hopstop-live-app-crowdsourced-transit

Zimmerman, J., Tomasic, A., Garrod, C., Yoo, D.,  Hiruncharoenvate, C., Aziz,  R.,  Thiruvengadam, N.R., Huang, Y. and Steinfeld, A. (2011), "Field trial of Tiramisu: crowd-sourcing bus arrival times to spur co-design," in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, Vancouver, BC, Canada, pp. 1677-1686