London’s transport watchdog, London TravelWatch, recently singled out TfL Go for failing to provide the information needed to plan a journey which avoids crowded stations and busy trains.
This criticism is not new and nor is it exclusive to TfL Go, however. Four other apps were also rated as inadequate, and passengers have for far too long had to roll the dice on which train to catch and stations to avoid in the absence of good quality information. It’s an age-old problem and a consequence of the transport industry relying on the same methods to predict busyness levels.
Reliance on historic data
All transport information providers trying to paint a picture of busyness rely on historic data. This includes Google Maps, who were found to be more popular amongst Londoners than TfL Go.
Historic data has never been a particularly accurate way of predicting busyness. The pandemic, and its impact on traveller behaviour and patterns, has rendered it potentially misleading. For instance, hybrid working is here to stay, and nobody knows what long-term impact it will have on either daily population movement or the density of public transport.
We need to stop looking to the past and instead look to the future if we’re to improve the way we predict busyness on public transport. This can be achieved by using a data-led technology solution that analyses passenger’s real intent to travel.
Tracking future intent is a fundamentally different approach to anything used by the transport industry previously but is something that has long been used in other industries. For instance, online retailers analyse ‘purchase intent’ to do everything from price management through to on-site targeting.
My team have been working with the University of Birmingham amongst others to produce data that is unique – it now powers busyness alerts for passengers using the National Rail Enquiries Alert Me service. We use it to tell passengers the ‘relative’ levels of busyness – for instance, your train is going to be busier than the next one.
We need to stop looking to the past and instead look to the future if we’re to improve the way we predict busyness on public transport
If we’re to get to a point where we can, for example, tell passengers if their train is likely to be over 75 per cent full or not, we’ll need to calibrate our predictions against reality. So, we’ve developed an artificial intelligence (AI) model in partnership with the University of Birmingham to do exactly that, but for it to work properly it needs feeding with historic data showing how busy trains were. The downside to this? The data is often withheld by industry because of perceived concerns over privacy and proprietary rights.
Although the rail industry has developed a framework on how to handle this data confidentially, it has failed to find a way to share it ethically and securely for the purposes of improving passenger information. A solution needs to be found if we’re to get better quality information in the hands of passengers.
Transport data, including signalling infrastructure and passenger counts, doesn’t need to be made open by default – it only needs to be accessible in a controlled and legally sustainable way. Any solution needs to be transparent and nor should it be commercially led or managed. It must have independent governance, accountability and regulation and have a structure that respects the legal rights of data suppliers, while encouraging the distribution of data in a commercially sustainable and legally viable framework.
Academia and industry
Birmingham Centre for Railway Research and Education (BCRRE) launched the UK Rail Research Innovation Network (UKRRIN) Data Platform last year, and it could provide the answer to this problem. It is a collaboration between academia and industry – engaged members include Network Rail, HS2 and Siemens Mobility – and has become the largest single source of industry data, both historic and real-time.
Although the UKRINN data platform is currently used for research purposes only – it helps unlock opportunities for organisations unsure what they can do with their data – it could be further developed, and licenses be put in place, so the required data can be shared ethically with commercial partners wanting to deliver services designed to improve the passenger experience. It could be the mechanism used by industry and government to deliver on the objectives set out by the recently launched Rail Data Marketplace.
Tracking future intent is a fundamentally different approach to anything used by the transport industry previously but has long been used in other industries
Another consideration is that many in the transport industry are at risk of letting perfect get in the way of good by insisting that predictions need to always be 100 per cent accurate. It’s unrealistic to think that we’ll ever be able to know in advance what exact passenger counts on trains will be, and nor is it necessary so long as predictions are being used to indicate how busy a train is likely to be or report delays.
If we’re serious about getting better quality crowding information into the hands of Londoners, we need to leave historic data in the past and replace it with data based on travel intent. If done properly, transport operators in London would know how busy their services are likely to be and be able to offer passengers incentives to take quieter routes at busy times, or simply let them know it’s going to be busier than usual and suggest another way home.
Let’s not accept the status quo as we return to work – the technology is there to improve our lives for the better. We should use it.