The science behind Grab

These days, Grab has remained the top choice among consumers who want their services fast and efficient


THE last few years could perhaps be the most revolutionary period for one of the most important segments for the Malaysian transportation business.

It was a time when the people were given a wonderful choice — to either stick with the traditional taxi for a certain price that has been determined and calibrated to the national standards, or “hitch” a ride with some strangers who would use their personal vehicles to make some extra money via a ride-hailing system that is also becoming a new norm globally.

The choice is made even more attractive when the new service is deemed to be priced lower than the traditional taxi rides.

Until earlier this year, two giants — Uber Technologies Inc and Grab — dominated the scene. Then, in a momentous corporate move, Grab bought into Uber to take control of the South-East Asian market.

These days, despite the number of smaller e-hailing companies operating all over the country, Grab has remained the top choice among consumers who want their services fast and efficient. After all, one is good to go with a smartphone in hand.

Still, has anyone ever stopped and wondered what the process is like from the time a passenger hits the “book” button on the ride-sharing app to the minute a driver arrives to honour the deal?

The good people at Grab would tell you that it is all down to data science and making the most out of it.

In Grab’s case, the company started off with mainly machines and as far as 2015, many “bread-and-butter problems” were encountered (and solved) before the management realised that being in the transport business, problems are actually solvable by non-machine learning techniques.

“We started off with only a few people. Today, we have about 47 teammates. We decided to organise ourselves into different teams and tackle as much as we could,” a Grab spokesperson said.

One of the team’s aims in using data science and artificial intelligence (AI) is to help tackle traffic initiatives, while working closely with academia.

“We have already started working with the governments in Singapore, the Philippines and Malaysia to better understand traffic conditions and help relieve congestion.

“Research was done on how we can apply AI into ride-hailing, from the time a person turns the app on,” he said.

Grab uses data science to predict the point of interest of a person; the estimated time of arrival of the driver; and how long the journey might take.

“We don’t necessarily send the closest car to a passenger. A lot of factors are considered before a driver gets sent the information,” he said. He added that data science is used to calculate trip fares as well.

“Grab uses something called a dynamic pricing system — where the price of the trip is influenced by the timing and the availability of drivers,” he said.

Despite the misconception about Grab’s pricing, he explained that it is very much like operating a marketplace, where the supply and demand chain reaction would be the main agenda.

“Based on economics, if there is a willing seller, there is a willing buyer — two people who see value in a product. The sole purpose is to match drivers to passengers,” he said.

While rewards and promotions are sent out by targeting a cluster of people according to their behaviour, or people who perhaps travel the same time every day, Grab said it no longer does blanket promos as there were plenty of inefficiencies.

Some people who received promotion codes were not using them, thus making it harder for Grab to evaluate the numbers at the end of the day.

Other than using AI for passengers, it is also used to predict the behaviours of its own drivers.

For instance, Grab can predict when drivers might start driving or which areas they are fond of circling.

“We try and match the behaviours of drivers, so we also give them incentives to entice them to accept bookings,” he said.

Grab also uses a heat map to identify where the high demand pick-ups are at — and so in letting the drivers know, they can conduct the matching easier.

“The ‘My Destination’ is a new feature, where drivers are able to enter a destination they are headed to and may want to do rides at, so they earn while on the way to a destination. It could be their home or maybe the office,” he said.

With team members watching traffic data, Grab is then able to notify its drivers of areas that are heavily congested or involve a breakdown of a vehicle.

“There was an experience where a train had broken down. Using this data, we were able to inform drivers that there may be many bookings being made in that particular area and so we were able to clear out the place fast enough,” he said.

AI has, and will, continued to evolve into the future — where combined with data and an understanding of human behaviour, it can help people develop patterns that will become the most useful.

“We want to disrupt technology, but not disrupt people’s lives in a negative way,” the Grab spokesperson said.