Case Study: Sequencing Supply and Demand in a Ride-Hailing Launch
Sector: Ride-hailing / passenger transport Markets: A multi-city expansion into new markets Engagement period: Confidential Scope: Driver and rider acquisition, city-level campaign architecture, measurement rebuild
Client identity is confidential.
The situation
The platform was expanding into a wave of new cities and using a single national campaign structure to do it.
The reported numbers looked acceptable. Cost per install was within target, install volume was growing, and the marketing dashboard showed a healthy trajectory.
The business numbers did not match. Day-30 retention was weak, a substantial share of installs never completed a first trip, and in several of the new cities rider complaints about wait times were rising while acquisition spend continued.
The national campaign structure was the cause. Optimising nationally, the platform concentrated spend wherever conversion was cheapest — which was consistently the cities where the service was weakest, because those markets had the least competition for ad inventory and the lowest cost per install. The system was reliably buying the least valuable customers available and reporting it as efficiency.
What the audit found
Supply and demand were funded from one budget. Driver acquisition and rider acquisition ran as a single "growth" line with no separate targets. When rider campaigns performed well on cost per install, budget shifted toward them automatically — in cities where driver density was already insufficient.
Measurement stopped at the install. No completed-trip data returned to the ad platforms. Automated bidding optimised toward whoever installed most readily, which is systematically not the population that completes trips.
Driver acquisition was measured on signups. Cost per active driver after four weeks was never calculated separately from cost per signup. When it was, the gap between the two figures was substantial.
No dayparting. Rider campaigns ran continuously, including hours when supply could not serve the demand generated. Failed requests during those windows produced churn among newly acquired riders — the most expensive customers to lose.
What we changed
1. Split supply and demand into separate campaigns with separate budgets and separate accountability.
Driver acquisition was rebuilt as a recruitment problem: different creative, different placements, and a metric of cost per driver active at week four rather than cost per signup.
2. Rebuilt campaign architecture to city level, and to zone level in the largest cities.
Each city received its own budget, its own supply–demand assessment, and its own pause authority. Cities where wait times exceeded an acceptable threshold had rider spend paused automatically until supply recovered — regardless of how efficient the acquisition looked.
3. Established the sequencing rule.
Supply leads demand, geographically. A city or zone does not receive rider acquisition budget until driver density passes a defined threshold and average wait time falls within an acceptable range. This slowed the launch schedule. It also stopped the platform paying to give first-time riders a bad experience.
4. Connected operational data to the ad platforms.
First completed trip, and subsequent trips, fed back as conversion events. Bidding shifted from install optimisation to completed-trip optimisation. Reported cost per acquisition rose immediately, because the platform was now paying for something real.
5. Applied dayparting to demand campaigns, aligned to actual supply coverage by hour and by zone.
6. Localised creative to city level.
Route names, landmarks, local price references. A city-specific creative library replaced the single national asset set, giving each market advertising that reflected its own streets, fares and reference points.
Results
The most visible change was that cost per install rose. That was expected and intended: the previous structure was cheap precisely because it was buying installs from people who never became customers. Paying more to acquire someone who actually completes a trip is a better outcome than paying less for someone who never does.
Retention improved because riders were no longer being introduced to the service in cities where it could not yet serve them well. Sequencing supply ahead of demand meant that a rider's first experience matched what the marketing had promised, rather than arriving to long waits in an under-supplied zone.
At the driver side, recruitment budget moved away from cities where signups were cheap but retention was poor, and toward a smaller number of markets where drivers stayed active. Over time this produced a set of cities that were structurally healthier — better matched supply and demand, lower churn on both sides, and a materially smaller share of acquisition spend wasted on customers and drivers who were never going to stay.
What transferred
Supply before demand, always, at zone granularity. Demand acquired ahead of supply destroys more value than it creates, and it does so invisibly — the acquisition metrics look fine while retention degrades.
Install is a step, not an outcome. Any marketplace measuring to installs is optimising toward the least committed population available.
National campaigns in a local business are structurally wrong. They concentrate spend where the service is weakest, because that is where the media is cheapest.
Further analysis of transport and mobility acquisition is published in publications, with the wider practice under business units.
To discuss a mobility acquisition programme, contact us.
