Two-Wheeler Last-Mile: The Analytics SEA’s Motorbike Fleets Actually Need in 2026

The Last Mile That Runs on Two Wheels

In most of Southeast Asia, the last mile arrives on two wheels. Motorbikes and scooters, ridden by employed and gig riders, carry the overwhelming share of parcels and food orders through dense cities and narrow lanes where a van cannot easily go. Yet most of the route-optimization and fleet-analytics platforms these operations run were designed around a four-wheel world: vans and trucks moving through Western road networks. The assumptions are baked so deep they are invisible, and they quietly produce wrong answers.

The result is a mismatch a VP of Supply Chain pays for without always seeing it. ETAs are off because the travel-time model thinks like a car. Routes are suboptimal because the tool does not know a bike can take the alley. Capacity is mis-assigned because the system reasons about van loads. Cost visibility is muddy because per-drop economics differ by vehicle type. None of these failures is dramatic on its own; together they erode service and margin across millions of deliveries.

More than 85% of all households in Thailand, Vietnam, and Indonesia own at least one motorized two-wheeler, and the ASEAN region accounts for roughly a quarter of global two-wheeler sales.

This piece sets out the analytics a two-wheeler-first fleet actually needs, and why treating a motorbike as a slow car is the root error behind most of them. It is written for the Southeast Asian supply chain leader who runs a two-wheeler network on tools that quietly assume four.

Why Four-Wheel Analytics Quietly Fail on Two Wheels

The core problem is a single hidden assumption: that a delivery vehicle is a car or van, and a motorbike is just a smaller, slower version of one. It is not. A two-wheeler is a different class of vehicle with different physics, access, capacity, and exposure, and every analytics output built on the car assumption inherits the error.

Also Read: AI Dispatch for Q-Commerce Rider Productivity in ID and PH

Consider how the assumption propagates. A travel-time model calibrated on car speeds will misjudge how a motorbike moves through congestion, because bikes filter through traffic that stops cars. A routing engine that only knows road-network links will miss the lanes and shortcuts a bike can use. A capacity model built for pallets and van loads will reason poorly about what fits on a bike. A cost model averaged across a fleet will hide the very different economics of a two-wheeler drop. Each is a small distortion at the parameter level, but they compound into ETAs the customer does not trust and routes the rider quietly ignores.

Southeast Asia’s digital economy surpassed $300 billion in GMV in 2025, driving a parcel base that independent estimates put in the tens of billions of deliveries a year.

The tell is when riders routinely override the plan, because they know things the system does not. That is not a rider problem; it is a signal that the analytics are modeling the wrong vehicle. Fixing it means treating the two-wheeler as a first-class vehicle type, not an afterthought scaled down from a van.

A Motorbike is Not a Slow Car

The most fundamental correction is the travel-time model. A car-based model estimates journey time from road-network speeds and congestion as a car experiences it. A motorbike experiences the same roads differently: it filters between lanes, keeps moving where cars are stopped, and is slowed by different things. Scaling a car’s estimate up or down by a fixed factor does not capture this; the relationship is not linear.

Drivers in Manila lost 143 hours (nearly six full days) stuck in traffic in 2025, with the Philippines the most congested country in Asia — exactly the conditions where a motorbike keeps moving and a car does not.

Accurate two-wheeler analytics use travel-time models built from two-wheeler movement data, so an ETA reflects how a bike actually crosses the city, not how a car would. This matters because the ETA is the foundation everything else rests on. A routing engine optimizing against wrong travel times produces wrong routes; a promise made on a wrong ETA breaks. In a two-wheeler-first network, getting bike travel times right is the single highest-leverage analytics correction, because every downstream decision depends on it.

Small Loads Rewrite the Optimization

A two-wheeler carries a fraction of a van’s load, and that constraint changes the optimization problem, not just its inputs. With a small carrying capacity, the question of what can be assigned to a rider at all becomes central: a bulky or heavy item may simply not be bike-deliverable, and a batch that fits a van is impossible on two wheels.

Analytics built for larger vehicles tend to over-assign, packing routes that look efficient on paper but cannot physically be carried. A two-wheeler-aware system treats load capacity as a hard constraint per rider, and reasons about batching within it, which produces smaller, more frequent trips and a different network shape.

For a VP of Supply Chain, this is why simply porting a van-based plan onto a bike fleet fails. The capacity constraint is not a detail to tune; it reshapes how work is bundled and dispatched, and the analytics have to model it from the start rather than discover it when riders cannot carry the load.

Also Read: Live-Stream Surge: Why Social Commerce Breaks SEA Fulfillment

Access and Parking Change the Map

Two-wheelers reach places four-wheel vehicles cannot. In the dense urban cores, kampungs, and narrow lanes common across Southeast Asia, a bike can navigate and park where a van has to stop far away and complete the delivery on foot. This changes the effective map: the set of directly reachable addresses, and the real last hundred meters, differ by vehicle type.

Analytics that model only the car-navigable road network miss this entirely. They either treat addresses unreachable by van as problems, or they ignore the bike’s advantage and route as though every vehicle faces the same access. Neither reflects reality.

Two-wheeler-aware analytics account for where a bike can actually go and how it completes the final approach, which changes both routing and realistic time per stop. For an operations leader, this is a large source of the gap between planned and actual performance, because the last hundred meters, invisible in a car-centric model, is where a meaningful share of two-wheeler delivery time is really spent.

Weather and Safety Are First-Class Variables

A van is an enclosed box; a motorbike is a person exposed to the elements. That difference makes weather and rider safety core analytics inputs for a two-wheeler fleet, not the footnotes they are in four-wheel systems. Rain slows a bike more than a car, makes some routes unsafe, and changes rider availability, and it does so in real time, not only across a season.

Analytics that treat weather as a minor adjustment will misjudge ETAs and over-commit riders in conditions where they should slow down or stop. A two-wheeler-aware system factors current conditions into travel times, allocation, and what it is reasonable to promise, and it treats rider safety as a constraint rather than an externality.

For a VP of Supply Chain, this is both an operational and a duty-of-care point. In a two-wheeler-first region, a stack that cannot reason about weather and safety at the level of the individual rider is missing a variable that materially drives both performance and risk, every rainy afternoon.

The Economics Live at the Drop Level

Two-wheeler economics are different, and managing a two-wheeler network well requires seeing them at the level of the individual drop. A bike is cheap per delivery in a dense zone but capacity-limited, so its cost-effectiveness depends heavily on density, batching, and the mix of stops in a way that a fleet-level average hides.

Analytics that report cost per delivery as a blended fleet number cannot tell a VP where the two-wheeler network is actually profitable and where it is not. Drop-level economics, broken out by vehicle type, zone, and time, are what reveal whether bikes are the right mode for a given slice of demand or whether it should shift to another vehicle.

This matters most in mixed fleets, where two-wheelers, cars, and vans coexist. The decision of which mode handles which demand is an economic one, and it can only be made well with visibility into per-drop cost by vehicle type. For a two-wheeler-heavy operation, that visibility is the difference between managing the network and guessing at it.

How This Works in Practice

Treating the two-wheeler as a first-class vehicle type is something an agentic platform such as Locus is built to do. Locus models vehicle-specific characteristics within the 250+ real-world constraints its engine enforces, so a motorbike’s travel-time profile, load capacity, access, and cost are represented distinctly rather than inherited from a van. Its Dispatch and Capacity agents assign and re-optimize work with those two-wheeler realities as inputs, through a continuous Sense-Decide-Execute-Learn loop that folds in live conditions such as weather.

Because vehicle type is modeled rather than assumed, the same platform can orchestrate a mixed fleet of two-wheelers, cars, and vans as one network, routing each slice of demand to the mode whose economics and access fit best. This runs at enterprise scale: 1.5B+ deliveries optimized for 360+ enterprise customers across 30+ countries, at 99.99% uptime. In one anonymized deployment, a Fortune 50 enterprise running 4,500+ drivers lifted its delivery execution rate from 75% to 92% through continuous, constraint-aware optimization, the kind of vehicle-aware decisioning a two-wheeler-first network depends on.

Also Read: How Does Locus Help Reduce Cost Per Delivery for CPG Distributors?

What This Means for a VP of Supply Chain

The practical test is simple: does your analytics stack model the motorbike, or does it tolerate it as a slow car? In a region where the last mile is overwhelmingly two-wheeled, that distinction decides whether your ETAs, routes, capacity plans, and cost reports are built on reality or on an imported assumption.

The audit is worth doing lane by lane. Where riders routinely override the plan, where ETAs miss, where per-drop cost is invisible, you are usually looking at four-wheel analytics applied to a two-wheel fleet. Correcting it does not require abandoning the tools so much as insisting they treat the two-wheeler as the first-class vehicle it already is on the road. In a two-wheeler-first market, the analytics should be two-wheeler-first too.

Learn more, visit locus.sh.