While announcements such as Network Rail’s plan to automate signaling through digital technology will help fix the age old problem of trains not running on time, data technologies could make that other customer bugbear of unpredictably priced and often over-priced rail tickets a thing of the past.
By the 2030s, almost three-quarters of UK train journeys will be controlled by automatic signalling under a project to use digital technology to replace analogue systems, saving billions of pounds. Network Rail said more than half of the country’s analogue signalling systems would need to be replaced over the next 15 years, and that a like-for-like update would cost £20bn, with no real benefit for customers.
There is good reason to do this. UK train passenger numbers have doubled over the past 20 years but the railway infrastructure, which is largely Victorian times, is struggling to cope.
But what about ticketing? Whenever I decide to use a train I always end up being surprised at just how much I end up paying. My disappointment is always magnified by some clever clogs who tells me how little they paid. According to Trainline, which sells tickets via its mobile app and website on behalf of operators across Europe, rail passengers can save an average of 49% on prices if they pay when they first search for the journey. For example, an advanced single ticket for a trip from London Euston to Manchester Piccadilly costs £32 around 80 days before departure but rises to £87 two days before.
It is a journey from A to B and then back to A. It hasn’t even been Rocket science since George Stephenson launched the first steam locomotive with that very name. How hard is it to make getting the best ticket price easy?
Well not that easy it seems, which is why Trainline has its data scientists on the case.
I spoke to Fergus Weldon, the ticketing company’s director of data science last week. Trainline sells coach and train tickets in 36 countries with 60 million visits every month on average.
It seems that the latest data technologies and techniques are being applied to improve rail travel for travelers as well trough better pricing. Which I am sure will attract more users.
Weldon leads a team of 50 focused on data at Trainline. The company also has a few hundred software engineers
Weldon told me there is a wealth of data on the company’s systems that can be used to improve customer services. “We have a large pool of data to drive innovation for customers. We try and derive and understand the potential value in data sets that can help customers on their daily commute or on their holiday journey.” This can include giving them insight into when is the best time to buy a ticket. The company offers this service through its Price Prediction tool to customers in the UK.
“We need to understand what price looks like over time. Because so many people use our platforms it allows us to build up a really clear picture of how price changes up to the day of departure. From this we can build algorithms that help us predict when the ticket price you see is likely to increase,” adds Weldon.
The company has a large and developed data infrastructure that captures all the data such as search data, Weldon says. “We capture what journeys people search tickets for as well as the results of the searches. We then make it available to our data scientists in an easy to work with format which allows them to apply deep learning frameworks such as TensorFlow on top of the data to try to make even better price predictions for customers.”
The company is currently working on a project to roll out its Price Predictions service, which is currently only available in the UK, it in France and further afield.
The company is also developing technology to catch timely and contextual messaging about train service disruption. It uses twitter and it can read messages from the train companies using natural language processing.
It also has a cloud based data gateway that provides users of services such as websites and apps the ability connect to it.
This is not all the company has in the cloud. In fact Trainline moved fully to the cloud three years ago from its on-premise IT infrastructure, with everything is based in AWS. “We like this because it gives us autonomy and allows us to move a quite a quick pace,” says Weldon.