Rail and Road are two different kind of shoes
Three out of four long-distance trains in Germany arrive on time. Nevertheless, we tend to have a strong negativity bias when it comes to judging the punctuality of mobility services. Much more likely do we remember the time we didn’t make it on time to an important meeting than the countless times we actually showed up punctually. Consequently, it is high priority for transport and mobility providers – be it logistic companies, train and tram operators or bus companies – to communicate more effectively to customers in times of schedule disruptions. Data science is empowering this new service culture and can help transportation companies to fully utilize all data potential and effectively leverage it for their operations.
When talking about transportation providers, there is conceptual difference between those providers, which operate on streets (trucks, busses, cars, etc.) and those which operate on a fixed network of appropriated routes (mainly trains, trams, subway). The difference is of vital importance, as it has consequences on the predictive challenges and focus of the respective companies. While operators which heavily depend on rail, are – in a short-term perspective – concerned about acquiring (and communicating) reliable information about delays and the arrival of the next train, operators with a free-floating fleet are focused on generating dynamic routing and dispatching to avoid major schedule disruptions. We’ll outline, how current data science approaches differ along the lines of transportation mode.
[Leveraging prediction power for transportation services]
On the road: Dynamic routing and capacity utilization challenges
Any company in the business of transportation on streets – whether it be people or goods – is faced with a continuous challenge to dynamically optimize routing. In case of traffic jams, logistic companies need to reliably inform their recipient about expected delays. The same holds true for bus companies like Flixbus, which needs to dispatch backup solutions in case of major schedule disruptions to ensure continuous service. Utilizing all accessible information is important in this case, going beyond sheer reliance on Google Maps, traffic jam warnings and standard solutions. Companies mine Twitter, Facebook & Co. to make sense of vast unstructured information and form a holistic real-time assessment.
Going beyond optimization of real-time monitoring, companies like Movinga, which is in the platform business for moving services, aim to dynamically route moving trucks, so that routes are composed dynamically along two dimensions, considering the real-time situation on the street and the geographic location of order volume. This is to some degree part of the “sharing economy” vision, aiming to fully utilize assets. Car-sharing providers are experimenting with similar ideas, integrating navigation systems with potential passenger demands. Given the flexibility of street transportation, have long moved beyond monitoring and look utilize prediction and optimization to inspire new business models.
On rail: Prediction challenges in the face of network dependencies
Rail transport, trams, metro and long-distance trains do not have the freedom of modelling their route according to external circumstances. On the contrary, tied to fixed schedules and routes, there can be various complicated schedule disruptions. Train A arrives delay at train station B, which consequently block the entry of train B. Long-distance train C has to wait for connecting passengers, which backfires on the entire schedule. Network dependencies evoke complexity, which makes it oftentimes difficult to even predict an answer to a question like: When does the next train in direction Y leave?
Deutsche Bahn has now moved forward to display at least the next two trains arriving on any given track, which makes travellers more comfortable. Projects like “Smart Data for Mobility” (SD4M), which aims to extract value out of unstructured data (Twitter, Facebook, blogs, etc.) is a first step towards more predictive approaches, but real value creation is still hidden in a more basic understanding of the essential question: When does the next train to my chosen destination arrive? This challenge is real for urban transportation as well as long-distance transportation and comprehensive approaches towards the issue are still lacking. Extensive understanding of the internal operating model in needed to tackle prediction challenges in field. This makes it rather inaccessible for innovation compared to road traffic, which can be modelled and predicted quite reliably with external data sources.
From a data science perspective, rail and road transport remains two different kind of shoes.
To fully utilize the potential of data science for its operations and business, the “road” ahead for rail-dependent transport operators will focus on getting the basics in place, focussing on the data quality and prediction within the network.