Data science brings the smarts to Grab Maps

Mapping apps have become an essential part of everyday life, from helping people find the best route for their commute, to optimising business logistics. The mapping engine that powers Grab’s deliveries and transportation services makes decisions based on terabytes of data points. Building a platform that effortlessly derives simplicity from complexity is hardly a simple job; it’s core to a data scientist’s job.

We sat down with Victor Liang, Grab’s Head of Data Science for Geo Road and Edge Computing, who’s been part of driving the company’s data innovations for over six years. From ETA predictions to mapping, Victor is focused on making sense of the data and using it to refine efficiency and productivity for Grab’s delivery network.

Making data-driven decisions

While Grab’s driver-partners use the app to navigate pickups and drop-offs, Grab is also able to collect data such as traffic situations from their activity. This helps the backend system to derive a more accurate estimated time of arrival (ETA) for a ride or delivery.

“ETA is everywhere if you use our app,” Liang said. “This is where we interact with our users to manage expectations. ETA is also used internally for our decision-making.”

Liang and his team analyse large quantities of data on a daily basis, in search of trends or possible areas for improvement.

The data generated by users is then fed into various predictive models. Grab employs a data science model for every city that the company operates in. “We get input from real-time traffic, weather data, users, driver speeds, routes… The more features, the more accurate our model,” he explained. “This is our daily work, how we continuously improve the accuracy of the Grab application.”

The team uses the data to derive useful, actionable insights. “We need to have an accurate map, we need to have good routing… these are all the fundamentals.”

Data and the human touch

All this information that’s gathered is ultimately used to help the app communicate more effectively with users. “Being able to provide an ETA reduces anxiety for our passengers,” Liang said. The more accurate the data, the more relaxed a passenger will be because they’re confident they’ll arrive at the time specified on the app.

Data also helps Grab consider every aspect of each delivery-partner or order—including (but not limited to) the weight of the food, the proximity of the consumer, and the type of vehicle most suited to the order. For example, a cake would be better suited for a motorcycle than with someone on a bicycle, he said.

“If we see that an order is relatively light, it’ll favour a delivery-partner on foot or a bicycle. A car or motorcycle might take the same amount of time, but you would have to factor in parking time, which might be longer than the trip itself,” he said. “We have to treat [every use case] differently.”

14 patents and counting

A doctoral graduate in Computer Science from the Hong Kong Polytechnic University, Liang started at Grab when it was a much smaller start-up. Since 2016, he has been honing not only his own skills but also refining the technology that consumers have come to know.

Some of his current projects include ETA predictions, positioning, and mapping technologies, and edge computing–running machine learning models with mobile devices. “In my team, we mainly work on traffic-related projects, such as ETA predictions, real-time traffic, routings, and so on,” he said.

Inventor is not a title we often see these days, but with 14 patent applications to his name, Liang is living up to the name. “When I joined Grab… we were less than 10 mappers. At that time—2016—we didn’t have tech [divisions]. So when we launched the Geo-tech teams, I was [one of] the pioneers.”