How Marketplace Chooses the Right GOJEK Driver For You
Have you ever wondered how GOJEK always manages to find you a driver when you need one? Do you notice how your driver is always just right around the corner to pick you up?
Well, it’s not magic; it’s GOJEK’s Marketplace team with a hefty dose of AI & technology.
The Marketplace team is responsible for making sure that its millions of customers are paired with the right drivers. This supply and demand matching is the core problem that the GOJEK platform tackles every day. And GOJEK is in the business of giving the best at what matters, including giving the best experience for its users.
From the moment the user clicks the “order” button to the second their driver shows up at their door, we want them to feel like it’s a match made in GOJEK heaven.
But what does it mean to be paired with the right driver?
“People want to get matched to a driver that is nearby, and drivers want the same. But the definition of ‘nearby’ changes based on the type of product and the time of day,” Andrew Brinson, Senior Vice President at GOJEK’s Mobility Marketplace told me recently. “We’re matching millions of orders weekly across Indonesia and Southeast Asia, being off by a few minutes is huge!”
To make it even more challenging, customer preference also changes depending on the time of day and the type of service.
After work, you might be willing to wait 10–15 minutes for your GO-RIDE or GO-CAR driver to pick you up. But during the morning rush hour, 10 minutes might just be too long because you don’t want to be late for work.
The same goes for the type of service. It would be understandable to wait 25 to 30 minutes for your GO-FOOD order to arrive because the restaurant might need time to cook your favorite dish. And it’s okay to wait even more than one hour for the GO-SEND driver to deliver your package to your mom across the city. But waiting more than 30 minutes for a GO-RIDE is much too long!
The Marketplace team must also make sure that it manages the availability of drivers on the road effectively. “So for every minute the drivers spend online, we want to make sure they’re making the most money when they’re doing trips and completing orders,” Andrew said.
To aid in the matching of orders, the Mobility Marketplace also looks at pricing, customer & driver incentives, and system stability across our tech stack. The team has individuals across product engineering, data science, business intelligence, strategy & operations in each of GOJEK’s offices.
So how does the Marketplace team make sure you get the right driver?
Using technology and AI, GOJEK’s Marketplace team decides which drivers are matched with which customers. Remember, they have to take into account so many factors and must balance the needs of both drivers and customers.
Yes, this job is as challenging as it sounds. And this was one of the first problems that GOJEK’s Data Science team had to solve.
The team’s first solution was a stand-alone Machine Learning (ML) model and web service to rank drivers. It worked well for a while, until it didn’t. A single ML model has a difficult time in balancing multiple business objectives.
The team then built a new system for multi-objective allocation that combines ML models and real-time features with a high degree of flexibility and manual configuration. The new system is called Jaeger, named after the world-saving humanoid mechas in Pacific Rim. Jaeger is made possible thanks to the work of GOJEK’s Data Science Platform team, which have built many of the foundational blocks for the system.
Jaeger is a system of modular and easily extendible components. The system consists of four components, Feast, Lasso, Meister and Babel. Each component serves a different purpose, each just as crucial as the other.
After Jaeger’s launch, it was obvious just how big of an impact the new system had on GOJEK’s business. The smarter allocation decisions enabled 1 million additional completed trips within weeks of the system’s launch, with significant improvements seen in dispatch time, cancellation rates, driver utilization, and income, just to name a few.
And this is just the beginning for Jaeger. GOJEK will continue to develop new and existing models.
If you want to read more about Jaeger, click here.
Solving tomorrow’s problems today
The biggest challenge for the Marketplace team is GOJEK’s rapid growth. They have to make sure that the system they design at the start of each year is ready to handle something 2 to 3 times bigger at the end of the year.
“The biggest tradeoff is solving today’s problems while knowing that the solution for today’s problems is almost certainly not going to work in the future,” Andrew said. “The problems in the future might be much meaner, much bigger in the next 6 to 12 months!”
The scale of the Marketplace team’s work continues to grow by the second, and this year, it is estimated that GOJEK would process billions of orders from its different streams of services. It’s a scale that no other companies have even been able to achieve.
To tackle future problems, Mobility Marketplace leverages automation — algorithms & AI to make decisions. Keeping these at the core of our product design, Marketplace can build flexible systems able to handle billions of annual orders.
While Andrew is satisfied with the work that his team has done so far, he acknowledged that there was much more work to do, such as improving services in GO-SEND and GO-SHOP to make the marketplace much more efficient.
If you think you have what it takes to solve problems at scale, GOJEK is your playground. GOJEK will continue to grow and its problems will only become more complex. Click here if you think you’re up for the challenge!
—
So, what do you say? Does juggling GOJEK’s supply and demand sound like a challenge you’re up for? Click here to ignite your power with GOJEK and help us keep the world in balance!
Shout out to Peter Richens for writing this article on how GOJEK uses Machine Learning to match drivers and riders.
Illustration by Athiyah Alatas
