Why Matatu Saccos Lose Revenue on Empty Return Trips (And How to Fix It)

Nairobi has always moved fast. The city's matatu network — over 100,000 vehicles serving millions of daily riders — is one of the most responsive, adaptive transit systems on the continent. It evolved without a master plan, shaped entirely by demand, competition, and the ingenuity of the operators running it.

But fast doesn't always mean efficient. And adaptive doesn't always mean data-driven.

Over the past year, we've been quietly collecting GPS telemetry across 340 tracked vehicles spanning twelve partner saccos. What follows is what that data actually shows — not what we expected to find, but what the numbers say.

Finding 1: Peak Congestion Is More Predictable Than Operators Think

Conventional wisdom in the matatu industry treats traffic as random — an external condition that drivers navigate by feel. Our data tells a different story. Congestion on Nairobi's major commuter corridors follows patterns accurate enough to model with high confidence several hours in advance.

Thika Road northbound between 5:30pm and 7:15pm on weekdays adds an average of 34 minutes to trip duration compared to the same run at midday. That number is consistent to within six minutes across 90% of the days we sampled. It's not random. It's predictable — and predictable congestion can be planned around.

Finding 2: Empty Return Trips Are the Single Biggest Efficiency Gap

We tracked vehicle utilisation across the full operating day for a subset of 80 vehicles. The result that stood out most starkly: the average vehicle runs its return leg at approximately 42% of its inbound passenger load.

This isn't purely a demand problem. Significant portions of the empty mileage occur on routes where demand exists on the return corridor — it just hasn't been matched to supply because operators have no visibility into where that demand is concentrated. A demand heatmap updated in real time would, in theory, allow drivers to make different decisions about staging, route deviation, and pickup points.

In aggregate, the vehicles we track cover an estimated 18% of their total daily kilometres with zero or near-zero passenger load. That's not a rounding error — it's a structural inefficiency hiding in plain sight.

Finding 3: Arrival Time Variance Drops Sharply With GPS Tracking

One of the clearest signals in the data is the relationship between GPS tracking and schedule adherence. Vehicles on routes where operators actively monitor live positions show meaningfully lower variance in arrival times at key stages compared to untracked routes operating on the same corridors.

The mechanism isn't complicated: when operators can see that a vehicle is running late, they can act — reassigning a nearby vehicle, alerting waiting passengers, or adjusting dispatch timing from the terminus. Visibility creates accountability, and accountability narrows variance.

Finding 4: Idle Time at Terminuses Is Significantly Underestimated

Industry estimates for idle time — the time a vehicle sits stationary at a terminus or stage waiting for a full load — typically run at around 15% of operating hours. Our telemetry data suggests the real figure, for the vehicles we track, is closer to 26%.

The gap between perception and reality matters because idle time is the most addressable inefficiency in the network. Unlike congestion, which operators cannot control, idle time is entirely within the system's influence. Dynamic dispatch — departing on a time schedule rather than a load threshold — is already standard practice in better-optimised transit networks globally. The data suggests Nairobi's network would benefit significantly from the same approach.

What the Data Points Toward

None of these findings require exotic interventions. They point toward a single underlying need: operators making decisions with real information rather than intuition.

The matatu network is already efficient by the standards of informal transit globally. With better data, it could be efficient by any standard. That's the work we're doing — and the numbers suggest the upside is substantial.

We'll be publishing quarterly data reports as our tracked fleet grows. If you're a sacco manager or transport researcher who'd like access to the underlying dataset, get in touch.