Grab Rolls Out ML Dispatch v3 Across Six SEA Markets — Driver Earnings Up 4%, Indonesia Drivers Cry Foul

Grab began rolling out its third-generation dispatch system across Singapore, Malaysia, Indonesia, Thailand, the Philippines and Vietnam on Monday. The company's own data, shared in a press briefing, shows driver earnings up an average of 4% across the markets in the first two weeks of the rollout. In Indonesia, however, the country's largest two-wheeler driver association issued a formal complaint to the Ministry of Transportation alleging that the new dispatch logic systematically favours four-wheeled vehicle drivers over motorcycle riders on cross-area trips.

The launch is operationally significant for the region's ride-hailing market. Grab is the dominant operator in five of the six markets in the rollout — Vietnam is the lone exception, where Gojek's Vietnamese unit holds a stronger share — and dispatch-algorithm changes affect the take-home earnings of an estimated 2.6 million drivers across the region. The 4% headline figure, if it holds across a full year, would represent roughly $480 million in additional driver income at current ride volumes.

What the v3 dispatch actually changes

The technical changes to the dispatch system, summarised from Grab's developer-blog explanation, fall into three categories. The first is supply forecasting: the system now uses a neural network to predict driver availability in each cell of the city grid over the next 90 minutes, rather than the prior heuristic that used 30 minutes. The second is demand forecasting: similarly extended from 30 to 90 minutes, with the network now incorporating weather, event calendars, and historical-pattern features. The third — and the one driving the driver complaints — is what Grab internally calls "trip-chain optimisation": the system increasingly prefers to assign trips that are likely to chain into a follow-up ride in the same area, reducing dead-mileage between trips.

The trip-chain optimisation is the contentious part. Four-wheeled vehicles benefit more from trip-chain matching than two-wheeled vehicles because they have higher fuel costs and longer trip durations, meaning dead-mileage hurts their earnings more on a percentage basis. The new logic, by trying to reduce dead-mileage system-wide, allocates more of the high-value cross-area trips to four-wheelers — which is what the Indonesian driver association noticed.

The Indonesia pushback

Asosiasi Driver Online (ADO), Indonesia's largest two-wheeler driver association, formally complained to the Ministry of Transportation on Tuesday. The association's specific claim is that motorcycle drivers in Jakarta have seen average earnings decline 2% under v3 — versus the 4% gain the system shows on average across all driver categories — while four-wheeled GrabCar drivers in Jakarta have seen earnings rise 7%.

The Indonesia case is consequential beyond Indonesia. The country has 1.4 million Grab drivers, making it the company's largest single market by driver count. The country also has the most-developed driver-advocacy infrastructure of any Grab market, with ADO regularly engaging the Ministry of Transportation and the Indonesian Parliament's Commission V on transport policy. A Ministry intervention against the algorithm would carry market-wide weight.

Singapore's PTC reviews the algorithm

Separately, Singapore's Public Transport Council confirmed on Tuesday that it has "asked Grab for a briefing" on the v3 dispatch algorithm. The PTC has algorithmic-transparency provisions in its 2024 ride-hailing regulations that require operators to disclose material changes to dispatch logic when those changes affect driver earnings.

Singapore's response is procedurally pre-defined and not specifically critical of Grab. But the timing — within 48 hours of the regional rollout — suggests the PTC is monitoring the same earnings-distribution data the Indonesian association raised. A formal PTC review of dispatch fairness has not happened at Grab before; this would be the first.

What it means for the broader ride-hailing market

The v3 dispatch system is the most-publicised algorithmic update Grab has shipped since the company went public in late 2021. The headline 4% driver earnings improvement is real and beneficial. The distribution of that improvement, however, is unequal — and the inequality maps onto driver categories that already had different earnings profiles, which is the political problem.

The broader pattern across ride-hailing markets globally is that dispatch-algorithm changes tend to produce wins-and-losses across driver subgroups even when the average improves. Uber has been through equivalent disputes in the US and the UK. Didi has been through them in China. The Grab case is the first major instance in Southeast Asia where the dispute is happening in a regulator-engaged environment with formal algorithmic-disclosure obligations on the operator.

The algorithm-transparency question

One of the underlying tensions in the Indonesia complaint is that motorcycle drivers — who collectively make up roughly 60% of Grab's Indonesia driver base — do not have a clear technical understanding of how the dispatch algorithm assigns work. Grab's developer-blog explanation runs to several pages of intermediate-technical detail; the typical motorcycle driver does not read English-language developer blogs and has no representative in the technical-policy discussion that produced the v3 changes.

The Indonesian Ministry of Transportation has hinted in past public statements that it would prefer to see algorithm-transparency rules that require operators to provide simplified, plain-language explanations of changes that affect driver earnings — analogous to Singapore's existing PTC requirements but with explicit accommodations for driver-level comprehensibility. The v3 incident may be the catalyst that pushes such rules to formal regulation.

The harder question is what algorithmic transparency actually means at scale. Grab's dispatch v3 has hundreds of decision factors, several of which interact non-linearly. The full algorithmic specification cannot be reduced to a one-page driver-friendly explainer without losing material accuracy. The realistic compromise is probably a summary of "principles" plus per-driver dashboards that show earnings impact attributable to the algorithm changes — both of which Grab has now offered to provide in response to the Indonesia complaint.

Driver-union historical context

Indonesia's two-wheeler driver activism has been the most-developed in the ASEAN ride-hailing market for nearly a decade. The country's labour ministry recognises ride-hailing drivers as a distinct category for collective-bargaining purposes — a status not afforded in most of Grab's other markets — and Indonesian driver associations have successfully pushed for minimum-fare floors, accident-insurance mandates and tip-transparency rules through earlier regulatory engagement.

The Indonesia driver-rights infrastructure exists because the country has the regional combination of high two-wheeler ride-hailing penetration (Jakarta alone has over 800,000 active Grab and Gojek drivers), strong organised-labour traditions, and a parliamentary structure where Commission V regularly addresses transport-labour issues. Singapore, by contrast, has roughly 60,000 ride-hailing drivers but no equivalent organised-driver-political infrastructure.

The Grab v3 dispatch dispute will likely play out primarily through Indonesia's existing infrastructure — Ministry of Transportation guidance, possible Commission V hearings, ADO negotiation with Grab. Singapore's PTC will follow its standard review process, which historically has produced minor adjustments rather than wholesale algorithmic redesign. The country-by-country regulatory variance is a structural challenge for Grab's regional-uniform-system approach to dispatch.

Operational performance — the early-data view

Beyond the headline 4% earnings figure, three other operational metrics have emerged from Grab's first two weeks of v3 data. First, average passenger wait time has dropped 11% versus v2 across the six markets — a clear customer-experience win driven by the better demand-supply forecasting. Second, the proportion of trips completed within 2 km of where the driver finished the prior trip has risen from 38% to 49% — the trip-chain logic working as designed. Third, driver hours-online have increased 6% week-over-week, suggesting drivers are accepting more work because the system is producing better-paying matches.

The first and third metrics are positive across all driver categories, including the motorcycle drivers in Indonesia whose advocacy is otherwise critical of the rollout. The second metric — the trip-chain gain — is what shifts the four-wheeler-favouring distribution and is the source of the Indonesia complaint. The data is consistent with both Grab's "average earnings improved" claim and the ADO's "motorcycle drivers worse off" claim being simultaneously true.

What Gojek does next

Gojek, Grab's largest regional competitor (now under Goto Group after the 2021 merger with Tokopedia), has historically positioned itself in Indonesia as the more driver-friendly platform. The v3 controversy gives Gojek an opening to recapture market share among motorcycle drivers if it can credibly argue that its dispatch logic remains fairer to two-wheelers. Gojek's senior leadership has not yet publicly commented on the Grab rollout, but industry observers expect a coordinated response within the next several weeks.

The competitive dynamic matters because Indonesia's ride-hailing market is the only major regional market where Grab is not the dominant operator by ride volume. A loss of motorcycle-driver supply to Gojek in Jakarta — even a small one — would meaningfully shift the market-share equation in Grab's largest single market by driver count. The Grab leadership team is aware of this exposure; the question is whether the v3 dispatch's overall benefits justify the localised political and competitive cost in Indonesia.

Sources

Grab's press briefing and developer blog provide the primary documentation. The Straits Times and Tech in Asia carried the regional-rollout lede. Reuters Southeast Asia provided the Indonesia driver-association response. The Singapore PTC announcement is from the council's official channel.