The alleys in South Jakarta are winding and narrow, sometimes just one meter wide, barely enough to fit a motorbike. A delivery rider has to squeeze past laundry racks and watch out for kids, while looking for errant house numbers to ensure an on-time delivery.
Across Southeast Asia, drivers for super-app GrabGrabFounded in 2012 and based in Singapore, Grab is a super-app that operates across Southeast Asia, offering ride-hailing, food delivery, and many other services.READ MORE navigate similar alleyways and streets everyday. Searching for a location that’s not properly mapped can cause delays, anger customers, and result in cancellations. There is a lot riding on the accuracy of Grab’s map, which the company has built with inputs from drivers and merchants in eight countries across the region.
“The streets here can be so complex and unpredictable,” Harsiniawati, a Grab delivery rider in Jakarta, told Rest of World. “They are sometimes missing entirely from the maps.” Harsiniawati, who goes by one name, often maps streets for Grab, with a camera mounted on her motorbike.
Like most businesses, the Singapore-based Grab initially relied on third-party navigation platforms such as Google Maps and Here. But as it expanded, it began getting more complaints from frustrated drivers who had trouble finding locations, Sriram Iyer, product head at Grab, told Rest of World.
“We asked for simple fixes, such as changing the location of a place because the merchant had moved, but even those would take months to get fixed,” he said. Grab realized that in the fast-changing cities and towns where it operated, “the existing providers were not going to solve it for us,” said Iyer.
Grab turned to its 5 million drivers to help map the streets and alleys they traversed every day on bikes, cars, and on foot. Drivers earned extra pay for mapping, and the resulting maps were more accurate than Google Maps or Here, also capturing colloquial place names and hyperlocal landmarks, Iyer said.
GrabMaps began as a tool for internal use in 2017, and became the app’s primary navigation tool in 2022. Today, GrabMaps’ data includes over 65 million addresses and points of interest, the geographical coordinates of restaurants, pharmacies, schools, gas stations, ATMs, tourist attractions, and more. Grab has plotted about 80% of the points of interest it needs, said Iyer.
Drivers initially used off-the-shelf cameras to take pictures. But after experimenting with various cameras, including GoPro, Grab decided to build its own camera to improve the quality of the images, according to Iyer. This allowed the company to have full control of the data, “so that we could use the data to solve problems throughout our systems. The investment has paid off from the self-sufficiency perspective,” he said.
The data from drivers, consumers, and merchants in more than 500 cities in Southeast Asia has helped add more than 800,000 kilometers of missing roads to OpenStreetMap, the open-source mapping platform, according to Grab. Grab drivers contribute about half a million map inputs every quarter. The data is refreshed every day, with real-time inputs on road closures, address changes, and other updates, said Iyer. The map’s accuracy helps drivers reduce the duration of their trips by about 90 seconds, allowing them to take more orders, he said.
Southeast Asia’s digital map market could grow to be worth more than $43 billion by 2030, according to industry estimates. Grab’s hyperlocal focus sets it apart from the more car-centric Google Maps and Here, Mohit Sharma, research analyst at Counterpoint, told Rest of World.
“Google and Here are only looking at major or neighborhood roads, not the smallest streets,” he said. “Southeast Asia is not the biggest market for them … Their main focus is the biggest markets like the U.S. or Europe.”
Grab is not the only company building its own maps. In China, Alibaba’s Amap and Baidu Maps are widely used — the latter’s clients include Tesla and Huawei. Indian ride-hailing platform Ola recently ditched Google Maps, saying it would instead use its own Ola Maps, built on OpenStreetMap. Meanwhile, ride-hailing company Bolt recently said it would use TomTom’s real-time traffic data to calculate journey times and map the fastest routes.
“People want something more local, more specific, at a cheaper price point,” said Iyer.
At the heart of GrabMaps is KartaCam, a camera that Grab designed and built for mapping. The new iteration, launched in 2024, can be mounted on the rider’s motorbike, which means special vehicles are not required for mapmaking. It can produce detailed images that capture lane markings, house numbers, the names of shops, and even their opening hours.
Grab recently launched KartaDashcam for cars, a front-facing camera that can be attached to a car’s rear-view mirror to record street conditions in real time. Around 500 KartaCams are in use now, with plans to deploy “tens of thousands” more, Iyer said.
In addition to the advanced cameras, Grab is able to improve the accuracy of its points of interest “because real trips and deliveries are taking place requiring accurate locations,” Jianggan Li, CEO of research firm Momentum Works, told Rest of World. This is “a very strong value proposition for enterprise customers.”
GrabMaps currently receives over 800 billion API requests in a month for services such as routing, navigation, and place searches. It became the data provider for Bing and Amazon Location Service in Southeast Asia last year. Address verification company Loqate said in October that it would use GrabMaps to offer “more accurate and localized address verification services” to its clients, including e-commerce firms.
The growing popularity of e-commerce, electric vehicles, and driver-assistance technologies in Southeast Asia present an opportunity for GrabMaps to offer solutions such as routing, and charging network and availability, according to Counterpoint Research. The KartaCam’s ability to detect potholes, for example, also makes GrabMaps a useful resource for government agencies, said Iyer.
-
KartaView app as seen from the phone display. -
There are currently around 500 KartaCams in use by Grab.
Grab uses country-specific machine learning models to reduce gaps in the mapping data. It uses AI to filter and improve its images, and its partnership with OpenAI will allow for greater automation, better data extraction from images, and faster updates, said Iyer.
But while its technology is innovative, mapping is capital-intensive, and Grab has a tough fight against bigger mapping companies, Sharma said. GrabMaps cannot hope for the success that Amap or Baidu have, since China’s tightly regulated market favors domestic mapping companies. “Southeast Asia is an open market. Therefore, achieving accuracy and local relevance … will require significant time and investment,” he said.
In Jakarta, Ahmad Fauzi is among more than a dozen Grab drivers who are trained to map with KartaCam. He organizes his own mapping schedule, and prefers it to food delivery, spending the bulk of his day on mapping before taking orders for a few hours.
“It’s like an adventure,” said Fauzi. “And I’m glad to improve the map for other drivers.”
