How did you get that Grab ride?

How did you get that Grab ride?

Head of engineering Ditesh Gathani says the act of calling for a ride on its platform may appear simple, but the computing behind that is actually “incredibly complex”.

SINGAPORE: “You already have the Grab app, what do you actually do at work?”

That’s a question Grab’s head of engineering Ditesh Gathani had to field, from his family no less, now that the Singapore-headquartered ride-hailing company is up and running in eight markets in the region.

But the computing and other processes behind matching a passenger with a taxi is actually an “incredibly complex” one, said Mr Gathani during an interview with Channel NewsAsia on Wednesday (Jan 31), during which he shed more light into how the company is using its large data trove to continually refine its services.

Grab's Ditesh Gathani
Grab's head of engineering Ditesh Gathani. (Photo: Grab)

IMPROVING JOB MATCHES

Grab has amassed an army of engineers, data scientists and economists to help answer the most basic of resource allocation question in transport: How to get a taxi driver to the place where a commuter is waiting for one?

It’s not an easy one, and the company has had its fair share of challenges, the engineering chief said.

For instance, four years ago, it worked on the premise that whenever a passenger makes a request for a taxi or GrabCar, the job will be advertised to all available drivers within a certain distance radius, say 2km, Mr Ditesh explained.

However, he recalled that taxi drivers were complaining that these job notifications were “annoying” as they weren’t actually near the passenger, or that it was difficult to get to the location due to road conditions such as traffic jams or being on a one-way street and having to make a big detour.

“This led to a lot of frustrations,” Mr Ditesh said.

To address this, the engineering team created machine learning models that helped to collect various factors related to job advertisements such as “time of day, destination of the job, fare and length of journey”. These datasets are then meshed and analysed, he explained.

“Now, at any point in time, the team can determine the probability of when a driver would pick a passenger up,” the engineer said, which reduced the frequency of job ads by two-thirds but increased drivers’ earnings (Across Southeast Asia, drivers earned 34 per cent per hour more than the average worker in the markets they are in, a Grab spokesperson said after).

It’s not just the drivers’ data that is taken into account though. Mr Ditesh said commuters’ data such as whether they are Platinum members of the ride-hailing company’s customer loyalty programme, GrabRewards, and the number of times they had cancelled ride bookings, are also taken into account when requests are made.

INNER WORKINGS OF DYNAMIC PRICING

The engineering chief also helped demystify some aspects of dynamic (or surge) pricing, which was introduced here about 10 months ago. He noted that before this feature, pricings were “flat” for drivers in Singapore but, since then, taxi drivers are on average earning 20 per cent more than previously.

On the origins of the feature, Mr Ditesh, who is from Malaysia, pointed out that there used to be a tipping function on its mobile app. So when it rained heavily and taxis were difficult to come by, passengers could indicate how much they intended to tip the driver to incentivise them to come.

The result was that drivers tend to wait for promises of tips, while passengers would go on “bidding wars” as they put up higher amounts of tips for their rides.

“This was dynamic pricing at play,” he said, “But we wanted to have a more refined process.”

Today, his team has created an algorithm that has demand and supply as the main parameters for deciding the fare pricing. The pricing mechanism is also such that fares do not grow exponentially if, for example, the distance travelled is far and price caps are introduced to make sure of this, the engineer explained.

It also implemented a price cap – S$100 in Singapore – to make sure that the prices quoted for JustGrab are “not ridiculous”, he added.

SECURING USER DATA

With big data, comes big responsibility.

The mangling of Uncle Ben’s oft-quoted phrase to Spider-Man aside, Mr Ditesh said the securing of information users have entrusted to the company is a key priority for Grab.

A simple example of this was when a passenger had emailed him, complete with credit card details, to ask if the person’s account is safe and has not been compromised, he shared. He said he had to instruct the IT team to make sure the financial details are redacted and purged from its email system.

Even when users call in to their call centres, the customer service operators can only access basic information and their access is logged for accountability, he revealed.

Asked if there have been attempts to hack into Grab’s systems, Mr Ditesh said: “It happens.”

That said, on a scale of P0 (“when we drop everything and try to fix the system”) and P4 (when it does not much harm), most of the security incidents fall on the latter end of the scale, while P0 events take place about “once or twice a year across its operations”, he added.

So even though it has internal security teams, Grab has engaged external consultants to conduct penetration testing to surface any gaps in their security posture, the engineer said.

The company also organised a public bug bounty programme last July, run by HackerOne, and offered rewards of up to US$10,000 for anyone able to identify vulnerabilities on its platform. 

Source: CNA/kk

Bookmark