The London bus is dead. Long live the London bus.
CityMapper is looking to revolutionise mass bus transport by using machine learning to create adaptive networks in London. Optimising for some, however, could mean everyone loses out.
Stroll across London Bridge on any given day and you’ll likely see an orderly queue of red buses crawling over the Thames river. This parade has gone on for generations; Transport for London (TfL), London’s statutory transport body, has been both operating and regulating the ‘only show in town’ for buses since 1933 [1]. Recognisable, cheap and multitudinous in the city centre, there’s little wonder that tourists and commuters alike flock onboard.
However, in the multi-billion pound economy of London transport, change is afoot. Buses, and the competition are smartening up – in more ways than one.
In February 2017, CityMapper – a transport technology company – launched SmartRide in London, promising an alternative to TfL. Half-bus, half-minivan, these unique vehicles have no fixed route or stops. Instead, they rely on a combination of machine learning, user input and human judgement to flexibly trace a route according to usage patterns and anticipated demand. Once requested to a location, customers step aboard and take their seats – identifiable by an allocated ‘Busmoji’ hovering on a screen by the seat [2].
Reading CityMapper’s press releases, you might be forgiven for thinking public buses already should be a thing of the past. In their own words, “the manner in which buses are allocated and run across cities is inefficient” and that “regulation makes it hard to be smart” [3]. The SmartRide solution arguably offers a neat fix, able to automatically adjust route and schedule to demand and thereby better meet customer needs. There is also a size-advantage: the minibuses employed on these flexible routes are up to 5m shorter than their double-decked red counterparts and therefore can squeeze into many more of the narrow and twisting roadways that make up central London [4].
This makes a worrisome issue for TfL, who are facing challenges both on and off the bus network. Bus passenger numbers have dropped for three consecutive years, and are down 6% from their highs in 2014/15, with broader customer satisfaction down 5% [5]. The Economist suggests stubbornly high traffic levels in the city centre are to blame, caused by large numbers of app-ride vehicles (e.g. Uber, Gett) and endless detours from roadworks which have in themselves reduced roadspace in favour of cyclists and pedestrians, compounding the issue [6].
So far the reaction to CityMapper SmartRide has been mixed. While promoted social media content has been positive, route and ridership growth has not been explosive [2]. CityMapper applied for only two ‘routes’ in London, and has been granted just one by TfL [7]. This draws into focus the awkward role the public body plays as both the main operator and regulator.
As a regulator, TfL needs to maintain access to transport for Londoners, in line with its responsibilities set by government [1]. This has created long-standing rules for service frequency, routes, and bus stop location; for example, a passenger in the outer suburbs will be no more than 640 metres from a stop [8]. In its other role as an operator, TfL has designed routes which support the economics of serving this regulation: some are profitable, and subsidise others which make a loss. But crucially, it must serve all these routes to in order to meet its responsibilities to the state. As private providers, such as CityMapper SmartRide, introduce adaptive routing into these geographies, they will naturally serve the most profitable, unsubsidised demand, drawing away valuable revenue from TfL’s star routes and leaving them with, in the classic phrase, “the young, the old, the sick, and the unemployed” – most often non-paying state-funded customers [5]. The net result is that the taxpayer ends up paying more for the same level of service. It’s already not looking rosy: last year London’s bus services required an operational subsidy of £752m [5].
A solution requires both parties to get onboard. Most immediately, both TfL and CityMapper can benefit from engaging closer on route design, leveraging the data each organisation has on passenger demand and bus utilisation. In the underserved neighbourhoods of Eastern and outer London, CityMapper should be given more scope to operate experimental routings which are beyond the resources of TfL. Looking longer-term, TfL’s red buses need to carefully shift into adaptive routing, doing so in a way which mitigates the effect on the most vulnerable in London.
The basic premise of demand-centric bus scheduling is sensible: it improves efficiency in an expensive, asset-heavy public service. However, several questions remain unanswered. Does machine learning in this context simply reinforce existing demand patterns? How do regulators protect against a slow drain of buses away from deprived areas with limited alternatives? And ultimately, what model offers the best outcome for Londoners? All aboard!
Word count: 738 words
Sources:
- Transport for London. (2018, November 11). Retrieved from https://en.wikipedia.org/wiki/Transport_for_London
- Citymapper Smartride. (n.d.). Retrieved from https://citymapper.com/smartride
- Citymapper. (2018, February 20). Bad Bus (Part 2/3) – Citymapper – Medium. Retrieved from https://medium.com/citymapper/bad-bus-part-2-3-e22f8824d3b8
- Double-decker bus. (2018, November 07). Retrieved from https://en.wikipedia.org/wiki/Double-decker_bus
- Transport for London Annual Report 2017/18. Retrieved from http://content.tfl.gov.uk/tfl-annual-report-and-statement-of-accounts-2017-18.pdf
- London’s buses are losing passengers. (2017, October 19). Retrieved from https://www.economist.com/britain/2017/10/19/londons-buses-are-losing-passengers
- Private companies want to replace public transport. Should we let them? (2018, March 22). Retrieved from https://www.theguardian.com/cities/2018/mar/29/public-transport-transit-private-companies-citymapper-uber-whim-smart-buses
- Bus Planning Literature Review – Research and Report by JRC Ltd. Retrieved from https://www.london.gov.uk/sites/default/files/bus_planning_literature_review_jrc_ltd.pdf
This is an interesting article about the interplay between a public good – transportation – and its regulator-cum-operator, and the private sector, as represented by CityMapper. To your point on the profitability of routes, what’s interesting here is that the demand for transport and the supply for transport are in some ways mutually reinforcing: Real estate prices tend to be higher in better-connected areas. However, traditionally, transport infrastructure has been much less flexible than real estate pricing effects. If transportation addresses this issue and becomes more malleable, conceivably it could allow a better equilibrium to develop between accessibility and affordability. As you’ve noted, however, there are many factors at play that will frustrate this effort.
Great article about transport in London! I really liked the way the author has built up the argument, it does remind me a lot about the big “Red Bus” of London. It does open my eyes also to the way transportation must evolve, not only in London, but in many congested cities world over.
The author has mentioned details of machine learning, but has not mentioned as to why this technology is a must, and how this will span out in the short and medium term. What key aspects of machine learning will be important, how will it play a greater role, for CityMapper Smart Ride. What could be some of the limits of the technology as well?
However, great article, that really made me learn about how machine learning could potentially shape the transportation industry.
This was an awesome read; I really appreciate you writing it. For me, this scenario serves as a perfect opportunity for government and the private sector to build a more collaborative relationship. To me, the government should limit itself to only servicing (likely at a loss) only areas that the private sector is unwilling to service (at a reasonable profit). To fund that subsidized service, it can either charge private firms a fee or source money from the general funds account.
It’s fascinating to hear how CityMapper is leveraging demand data to more efficiently transport people across London. I’m optimistic that solutions like this will reduce traffic congestion and minimize the number of empty seats in vehicles. I would be curious to learn more about how the cost of CityMapper’s bus service influences the demographics of the riders. Given that CityMapper’s service requires smartphone technology, I suspect that the demographics of the service skew younger and likely middle- to high-income. I therefore echo your concluding concerns. My fear is that privatizing public transportation will disproportionately impact individuals who live in underserved areas with fewer alternative transportation options. Given that public transportation is designed to be accessible to everyone, it’s important that we are not using historic demand as justification for discontinuing public transportation service to parts of cities where fewer users of CityMapper’s service live.
Fascinating article.
In my view, the competitiveness that will result from SmartRide’s entry into the market is bound to be highly beneficial for the population.
First, it will be an additional service, risking to die naturally if it fails to provide any valuable service to the community. Second, it is a collective means of transportation, hence advantageous in comparison to ride hailing companies. Third, it will put pressure on the regulator to watch out for trends and keep/improve the quality of the buses.
Therefore, I see several upsides and admire this project. In case it fails, I am certain that others will then come up with better ideas, as long as the regulator allows innovation to flow.
Fascinating look at disruption and machine learning in a traditionally backwards area! From a policy standpoint, it sounds like CityMapper and technology are scrambling the implicit public transportation bargain: government is granted monopolies on bus, subway, and light rail service in exchange for fair access, funding from progressive taxation, and sustainability of the market. Machine learning has given CityMapper the ability to provide a differentiated service good and cheap enough to draw users away from TfL. This is remarkable. But I think, however, that CityMapper should be forced to pay for the externalities it is imposing on public transit networks. The continued existence of public transit requires the cross-subsidization you mention. In unwinding it, CityMapper is shifting value from the public to itself (and to be fair, the people who use the adaptive bus). But, it sounds like Londoners have agreed to this bargain before. Citizens should rightfully ask CityMapper to cover some costs of access for those worst off.
Fascinating article that reminded me of a similar initiative in Singapore (https://www.beeline.sg/). Demand-centric service definitely can benefit a lot with sufficient scale, as the company gradually collect datapoints on where the most demanded routes are at different peak/non-peak hours. For one, I could imagine intuitively that there would be more services heading towards the central business districts in the morning and departing from there in the evening during the workdays. That said, you brought up an interesting point in how critical regulator plays a role in this business, as the bus operator would need to still maintain a minimal level of service. I cannot agree more that there would be a lot to gain if both companies work together, and might even go as far to suggest that the demand-centric company eventually fold in with the bus operator, so that there is less duplication of routes and a ‘masterplan’ on which route should have ‘traditional’ bus services versus ‘demand-centric’ services.
Fascinating article. I like how it brings up some of the key challenges that demand-driven for-profit transport services pose to traditional public transportation networks. While these new transport services like CityMapper in London or the former Bridj in Boston provide on-demand, high quality, and low-cost transport services to some, by compete away revenue from existing public transportation systems, they could have a significantly negative societal impact. Affordable, efficient public transportation is key to social mobility and is often cited as a major reason why social mobility is higher in Western Europe than in the US. Preserving high quality public transportation for all must be the mission of TfL even if it requires heavily regulating services like CityMapper.
The CityMapper flexible bus solution reminds me of the Harvard Shuttle. While public transport within Cambridge and specifically near the HBS campus is lackluster, the Harvard Shuttle system offers a convenient solution to students. Certainly, there are very common routes that students need to take (e.g., HBS to Harvard Square) and relying on the community government to provide a public transport solution may not be ideal. I think it’s reasonable to assume that the larger Harvard Shuttle buses travel along the most common routes while solutions such as the Harvard Evening Van allow students to specify their (more unique) routes when the larger buses are no longer running. Perhaps Evening Van routes taken most commonly then evolve into larger bus routes. In a similar vein, I think there would be tremendous value in community transportation agencies leveraging data from Uber/Lyft about top routes taken by users to develop new public transportation routes or modify existing ones to better serve customers (though there are incentives for them not to share this data).
Great article, hitting on the crux of the matter in my view: how to ensure that TfL buses are not left with the most unprofitable routes, making it even more expensive to the tax payer? Even though I am a leaning libertarian, I would support a more stringent approach in this situation; granting TfL monopoly is the only way to ensure the broader society can access affordable public transportation without requiring outrageous subsidies from the tax payers. Hence, I would advocate for TfL to purchase CityMapper services, and use machine learning to optimize its routes, providing a much better service to its customers.
CityMapper has the benefit of so much more data than TfL, and even if perhaps not in quantity if TfL is collecting all the OysterCard data, it is definitely not able to leverage as smartly as CityMapper. How scalable truly is CityMapper? London buses have a massive infrastructure that I am constantly impressed with, every bus is clean, efficient and on-time. It has its bus only lanes and buy in from Google and other apps that run live bus schedules and absolute transparency. There’s no way I can see CityMapper getting to that scale and efficiency. It seems almost a glorified Uber, and honestly while Uber is cheaper than cabbies in London, I still sometimes get on a bus rather than an Uber. It’s still sometimes more convenient because the stops are just there and go almost everywhere from anywhere. You can’t beat that!
This was a very interesting read to know how the transportation system is evolving in London. It was also interesting to see how the disruption in private sector (CityMapper) is causing a problem for the mass public as a whole by causing inefficiency in TfL.
A possibly radical idea I had was, “What if the city opened TfL to be run by a company, like a company (with restrictions on keeping the bus stops in some area, to protect the people)? Would that help foster competition and improve the efficiency of the city as a whole?” The idea came to my mind because that is basically what happened for postal service business in Japan. From late 1990s to 2000s, Japan Post, a massive 250 trillion dollar organization, became privatized to promote competition with transportation companies and help the organization become more efficient. As a result Japan Post has been able to hire industry experts in various fields, from logistics to wealth management, to improve its effectiveness. While collaboration between CityMapper and TfL is certainly ideal, I thought of opening up the operation of TfL to industry experts could be another way to advance the London transportation system.
This is a great essay on how eminent and initiate machine learning can be. The technology does not have to sit in a super lab in a college campus, it can be used in familiar ways to address even more familiar problems.
I’m wondering what will be Uber’s response to CityMapper, given that Uber already have the algorithms, data, and the experience. In any case, rejuvenating the public transportation system in a buzzing large metropolis will create meaningful benefit such as reducing time wasted in travel, reducing energy consumption and emission, and improving traffic congestion.
I wish CityMapper luck and look forward to more of these entrepreneurial problem solvers.