Trainline uses crowdsourced data to help commuters social distance on the return to the office

Profile picture for user Mark Samuels By Mark Samuels September 7, 2020
Summary:
Trainline CTO Mark Holt says his team worked hard during lockdown to develop new ways to keep travellers informed and safe.

Image of a woman wearing a mask covid-19
(Image by NickyPe from Pixabay )

Transport specialist Trainline is using crowdsourced data to power a real-time service that keeps people socially distanced as the return to the office gathers pace.

While many people are still working from home due to the coronavirus pandemic, increasing numbers of employers are reopening their headquarters following months of lockdown in the UK. The return to the office is likely to be an anxious time for workers - especially with the thought of travelling on jam-packed public transport, where maintaining social distance is a challenge.

That's where Trainline aims to use its data to help keep commuters informed and safe. The company has developed Crowd Alerts, a crowdsourced feature on the Trainline app that tells travellers when it's not possible to social distance on a specific train.

The company beta-launched Crowd Alerts at the beginning of July. Data from a two-week trial of the feature prior to launch showed that it was possible to social distance on 90% of trains. Trainline CTO Mark Holt says the rapid development and launch of the feature has been due to the great work of his 400-strong technology team:

It took six weeks end-to-end to ideate, design, build, run multiple rounds of customer testing, and then ship the product. And that's testament to the team's abilities to work remotely and engage with each other when they were all sitting at home.

Customers who want to use Crowd Alerts tap the ‘live tracker' button for a journey in Trainline's app. A pop-up message asks the customer to confirm whether social distancing is possible on their train by clicking a ‘thumbs up' or ‘thumbs down' button. If the customer clicks ‘thumbs down', they're asked to share whether they are in the front, middle or back of the train. With this information collected, Trainline's data-processing system sets to work:

We churn that data. This is important - and we took this product very seriously, because what we didn't want to do was create fear. We also didn't want to tell people that there was good social distancing if there wasn't. We just wanted to be really careful with the product conception. And so we churn the data quite a lot because we don't want people manipulating it.

Crowd alerts

The crowdsourced results are sent to other users of the app - an orange bubble indicates a busy train, while a blue one says a train is only partially crowded. The data-led system is designed to stop passengers from falsely reporting a busy carriage. The bubbles indicate how many commuters have shared the information and fresh feedback overrides previous reports. Trainline uses Apache Kafka, the open-source event-streaming platform, to process customer feedback and to ensure the real-time feed is as clean as possible:

We have an event-sourcing platform that supports the app. So every event that happens on the platform is spat out into a Kafka instance and then that's churned. We store all those instances in our data lake and we also turn them in real time into various data products. That's one of the joys of Kafka - you just simply create a new sync for the data and it starts to flow and then you churn it in the appropriate way. It's a very powerful piece of tech and we use it for a lot of tools.

One of these tools is BusyBot, which crowdsources data from passengers to provide real-time information on seat availability. Holt says his team used BusyBot as the backbone for the new Crowd Alerts system. Victoria Biggs, Trainline's vice president for brand and communications, came up with the "spark of inspiration" for adapting BusyBot. Holt's team then turned this idea into a practical reality:

The nice thing is that we already had the infrastructure in place to process busyness data. That meant we also had experience with crowdsourcing. We knew that 26,000 people were dropping data into BusyBot. That's cool - and that told us that we had enough critical mass to be able to do something around social distancing.

Holt says the key to research and development success is putting a specialist team on a project and letting them focus their efforts. His "big belief" as a manager is to stay out of the way and let smart people in his department make the application work:

You don't gain anything by adding a lot of governance on the back of a project. We set the direction and we let them let them run at it. So we started out with a product owner and some designers and some developers. Then we fleshed out the concept and almost the first thing we did was we showed Crowd Alerts to some customers.

Test, iterate, test again

Trainline's research and development effort always leans heavily on customer testing. In this instance, customers were asked whether they liked the Crowd Alerts feature and whether it would increase their confidence in their ability to travel. Holt says upwards of 85% of people said they liked the app and believed it would boost their confidence:

So we kept going. And as we kept iterating, we would update the prototype that we would show the customers, and we would then respond to their feedback. We knew where we were trying to get to and we were just charting the course to get there.

As more people start to travel by train again, Trainline intends to make the anonymised data it collects available to the rail industry so that operators can identify parts of the network that customers are reporting as busy. That brings benefits to customers, too:

The really interesting thing is that we're actually seeing great social distancing. We have a very cool real-time dashboard that shows us up-to-the minute information on where the routes are that have good social distancing. It's a crowdsourced product and it's super-adaptive. So as people come back, and start using rail, and they start providing more feedback, it makes the information more accurate. And all we want to do is give customers the ability to see that if there are two trains coming in - and if you don't get on this one, but you get on the next one - then you can social distance.