Riding on data for better efficiency
Grab's vice-president of engineering shares how the ride-hailing business uses real-time data analysis
Consumers can get a ride with just a few taps on their smartphones.
But for ride-hailing business Grab, it is a complex process of real-time data analysis that looks at the demand and supply in each 1km by 1km area, which driver to match the ride to and how much to price the ride - all done in a few milliseconds.
Speed is crucial or people may opt for another option, said Grab's vice-president of engineering Arul Kumaravel.
He was speaking to The New Paper on the sidelines of the Amazon Web Services (AWS) summit last week.
Grab's massive repository of data has allowed it to match supply to demand in real time.
For instance, it can send drivers to Tanah Merah Ferry Terminal based on ferry arrivals.
"Our system is able to pick up that there is a pattern of demand picking up at a certain time at this particular geo-location. And it turned out that every time the demand spiked was because of some ferry coming.
"This kind of real-time information is what our system is able to pick up, and then it can tell if there are drivers within a few kilometres, depending on the city," Mr Kumaravel said.
This kind of real-time data analysis is made possible by the AWS platform, which offers a suite of cloud computing services such as storage, analytics and Internet of Things.
Businesses can use any combination of the AWS services to build applications and run them on the AWS cloud.
For instance, the Elastic Load Balancing service allows Grab to align with peak hours and scale up on demand, and the analytics serviceallows Grab to react to demand on rainy days or when an MRT train breaks down.
As for Grab's dynamic fixed pricing algorithm, it is Economics 101 - prices are higher simply when there are more riders than cars on the road.
But things can swing the other way too, said Mr Kumaravel.
"If there is more supply, we encourage people to take Grab... So more people will say, 'Hey, instead of taking the MRT, it looks like the prices are reasonable, so we can take a Grab.'"
This "encouragement" is doled out through promotional codes, which are determined by the data that tracks how often one uses the Grab platform.
Mr Kumaravel said: "We use that as a mechanism to introduce a new product or service to our customers with the intention that once they use it, they will see the value we provide, and they will continue to use it."
GrabShare, for instance, has seen more than two million rides since its launch in December. This means that even though more people had started using the new service with a promotional ride, they are now repeating their GrabShare rides, he said.
But National University of Singapore's (NUS) Professor Ho Teck Hua pointed out that people are loyal to promotional codes, not the company giving them out.
The Tan Chin Tuan Centennial Professor and director of the Centre for Behavioural Economics at NUS told TNP: "If other systems are as easy to use, people may switch.
"People are using these codes, but they may not be loyal to a particular company.
"My hunch is that the cost of switching (to another app) is not high enough.
"Giving promotion incentives indirectly starts a price war among operators."
Riding on data for higher efficiency
It is always a quest for Grab to hit that sweet spot - where the demand from riders meets the supply of drivers at the right place, so that prices remain as low as possible for both the riders and company.
The ride-hailing platform tries to get as close as possible to this ideal by making sure that its drivers do not stay idle.
Its vice-president of engineering, Mr Arul Kumaravel, told The New Paper: "What we are doing is making sure that when (the drivers) are online with Grab, they are most efficiently used."
With a network of more than 50,000 drivers here, Grab tries to meet their needs by feeding them a variety of data - a combination of real-time and historical data that is stored on the Amazon Web Services cloud.
This includes gauging the odds a driver would bid for a job. And if a driver gets a ping about demand at Raffles Place but does not go there, the system will note it down and adapt, said Mr Kumaravel.
"It is not like we spam every driver with every (piece of) information. Our system personalises the information, knows which driver will respond to what kind of messages and knows what they prefer," he said.
New rides are also dispatched to drivers before their current one ends, making it an almost seamless transition from one trip to the next.
Grab driver Andrew Kee, 40, told TNP: "It is easier for us, because it means we can complete more trips. We make better use of our time on the roads because we are matched to more passengers and more accurately."
This state of mind - being immersed in completing trips and finding enjoyment in racking up the number of completed trips - should be maintained to keep drivers on the roads, said Dr Walter Theseira.
The Singapore University of Social Sciences economist told TNP: "Otherwise, the driver could get distracted and look for a break. That could reduce the driver's earnings and their availability."
National University of Singapore's Professor Ho Teck Hua added that for the drivers, seeing a ping for the next trip is almost like fulfilling an unfulfilled duty.
A more recent update for Grab drivers is an app feature that helps them track how far they are away from getting certain incentives.
This is something Grab drivers asked for, said Mr Kumaravel.
"This pushes drivers to work a little harder," said Prof Ho.
"On a bad day for drivers, they might take a shorter break to make up for the shortfall (in earnings) and meet their goals," he said.
Dr Theseira said: "The other side, however, is that studies have shown that once people meet the objective, they may stop driving for the day - so the objective has to be chosen carefully to balance incentivising people to meet it versus having them lose interest once they meet the objective."
At the end of the day, Grab hopes for its drivers to make their time spent on the platform count.
"From a driver's perspective, if you think about it, they are driving on a platform to make a living," said Mr Kumaravel.
"We need to make sure they are able to make enough income based on the hours they spend on our platform..."
- FOO JIE YING