Where Lagos commutes
After Conductor, a Lagos commuter carpooling app, relaunched, hundreds of new users told us where they live, where they work and how they travel. I turned those answers into a demand map and a supply check, and found the marketplace leaning the wrong way.
Situation
Conductor, a Lagos commuter carpooling app, relaunched in September 2026, and every new rider and driver told us where they live, where they work and how they travel.
Task
The growth lead asked where commuter demand sits, how regular it is, what it competes with, and whether driver supply is in balance with it.
Action
Snapped every home and workplace to a ~1 km grid, drew home-to-work flows, read the commute survey answers, and set rider sign-ups against the driver funnel.
Result
Demand converges on the business district ( of riders work there), two thirds commute five or more days a week, and drivers outnumber riders nearly two to one, so the growth case is for riders.
The setting
Conductor matches people who drive to work with people going the same way, for a fixed weekly fare. I work on it as the data consultant, and the growth lead asked me to read the relaunch cohort. When we relaunched, every new user, rider or driver, was asked for a home area and a work area. Many also answered a short survey on how many days they commute and how they travel each way.
That is sign-ups, of whom gave a commute. The map uses the home and work pins of the people in the relaunch cohort who placed both; the survey figures further down (days per week, travel mode) are on their own bases. Everything on this page is real, but aggregated: pins are snapped to a ~1 km grid and the map shows only percentages, so no id or exact location leaves the database. Internal and test accounts are excluded.
Where is this?
Lagos is Nigeria's commercial capital, a metro area of over 15 million people on the Atlantic coast of West Africa. Lagos Island, Ikoyi and Victoria Island, with Lekki Phase 1 just to the east, form the country's main business district, marked by the dashed box on the map. It is home to the Nigerian Exchange (the stock exchange), most major bank headquarters and many multinationals. For a UK reader, think of it as the City of London and Canary Wharf in one. Most workers live on the mainland and cross the Lagos Lagoon on a handful of bridges, so a few kilometres of destination collect commuters from across the whole region.
Home and work pins of people snap to a ~1 km grid, and each person is drawn at a fixed random spot inside their cell so the grid itself doesn't show. Nearby people group into clusters that merge as you zoom out. Clusters holding 2% or more of everyone in the current filter (All, Riders or Drivers) show that share; smaller groups are plain dots, and hovering any dot gives its share. Flows join ~5 km cells and show only home–work pairs with 3+ people, which covers of commuters; line width by share. The dashed box is the business district: Lagos Island, Ikoyi, Victoria Island and Lekki Phase 1. Hover for shares. Boundaries: geoBoundaries / GRID3 (CC BY 4.0). Tiles © Esri.
Origins scatter, destinations converge
Switch the map between Home and Work and the shape of the problem is obvious. Homes are spread thin: Ikorodu, Ikeja, Abule Egba, Ojodu Berger, Sangotedo, and over the state line into Ota and the Ogun side of the Lagos–Ibadan corridor. About of commuters live in Ogun State.
Workplaces are the opposite. of the people on the map work inside the business-district box (Lagos Island, Ikoyi, Victoria Island and Lekki Phase 1), and riders lean on it more than drivers: against . That is the tightest definition. Extending the box along the Lekki corridor to Chevron gives about , and the full Lagos Island and Eti-Osa local government areas (on to Ajah) about . I show the tightest one and say so. Ikeja is the clear second destination.
The grid itself shows the difference. The ten busiest ~1 km cells hold of workplaces but only of homes, and of people live in a cell they share with at most one other person.
For a carpool product this is good news with a catch. Many-to-one demand is easy to match at the destination end, since most cars are heading for the same place. The hard part is the pickup end, where a driver from Ikorodu needs riders near Ikorodu, not just riders going to Victoria Island.
Top named areas
Share of the people who gave a commute. Names are raw place text from search, so near-duplicates appear separately (see data notes).
Home
Work
Commuting is a weekly habit
The next question was whether this is occasional travel or a routine. It is a routine. of people who answered commute five or more days a week, and the single most common answer is exactly five. Only say one day.
Share of people who answered, n = . Darker bars are five or more days. Hover for counts.
That matters for the business model. Conductor sells a seat for the week, not a single ride. A rider who travels five days a week is worth five trips of matching effort once, and the same driver and route can serve them all week. Recurring demand also makes supply planning simpler: today's riders are a good forecast of next week's.
What we are really competing with
The commute-mode question asked how people travel to work and how they get home. The answer is not the one a ride-hailing pitch deck would assume. About half use public transport (danfo minibuses, BRT, keke) in each direction. Ride-hailing is a minority. So the real price competitor is a bus fare, not an Uber fare.
Most people do the same thing both ways: of those who answered both legs keep one mode. The interesting exceptions are in the "Morning vs return" view. Of morning ride-hailing users, close to half come home another way, most often by public transport or with a colleague. People pay for speed when they must arrive on time, then save on the way home. That is a price-sensitivity signal: a fixed weekly carpool fare has to beat that mixed spend, not a full week of ride-hailing.
Small sample on this question
Only people answered the mode question, against who gave a commute. A share of 14% here has a 95% margin of roughly ±6 points. I treat the order (public transport first, by a wide margin) as solid, and the smaller categories and switching rates as directional.
Supply was outrunning demand
The last question was balance. At sign-up, drivers outnumbered riders to one, and onboarding (finishing the basic profile on each side) shows the same ratio. My read is that the driver pitch, earn back your fuel money, is easy to say yes to, while riders need more convincing to change a daily habit.
But a driver who has signed up is not yet supply. Before a driver can publish a trip, Conductor checks phone, email, national ID (NIN), vehicle, licence and documents. The driver funnel shows where people drop out.
Riders and drivers since the relaunch
Riders vs drivers: share of the cohort at each stage, by side. "Onboarded" counts people who completed each side's basic profile, so a person can appear on both. Funnel: drivers reaching each verification step, as a share of driver sign-ups; labels show the drop from the previous main step. Vehicle sub-steps run in parallel, not in order. Hover for counts.
Only of driver sign-ups reach fully verified. The single biggest leak is right at the start: of drivers who verify their phone never verify their email. That is a step with no real friction except remembering to open an inbox, and it loses more people than the national ID and vehicle checks combined.
So there are two different problems. Raw interest is lopsided towards drivers, so the data points acquisition spend at riders on the strongest corridors (Ikorodu, Ikeja and the Lekki–Ajah axis into the Island) rather than at more driver sign-ups. Verified supply is a funnel problem, and the fix is product, not marketing.
What I'd change in the funnel
- Make email optional at sign-up. Ask for it later, when there is a reason (receipts, payouts). Phone is already verified; email adds little trust.
- Collect vehicle photos and documents in one sitting. The vehicle sub-steps lose people in parallel. A single guided capture flow with a progress bar is likely to hold more of them.
- Nudge by step, not by date. A driver stuck at email needs a different message from one stuck at licence upload. Each step is already tracked, so this is cheap to target.
Data notes
- Area names are raw search text. Home and work areas come from Google Places text, so the same place appears in several forms: "Ikorodu" and "Ikorodu garage market", "Victoria Island" and "Victoria Island, Lagos". I left them unmerged in the table rather than hand-pick a mapping; the map uses coordinates and is not affected.
- The mode question has a small base. of commuters answered it, because it was added to the flow later. Shares carry wide margins and may over-represent the more engaged users who finished the whole survey.
- Privacy by aggregation. Home and work pins snap to a ~1 km grid before they leave the database, with no ids and no exact locations, and the page shows only percentages, never counts. Every pin is kept, so the Home and Work views cover everyone. Flows join ~5 km cells and a home–work pair needs 3+ people to be drawn, so the flow lines over-represent the busiest corridors.
- The business district is a drawn box, stated openly. A workplace counts when its unrounded pin falls inside lat 6.42–6.47, lng 3.38–3.49 (Lagos Island, Ikoyi, Victoria Island and Lekki Phase 1). A box on to Chevron gives about ; the official Lagos Island and Eti-Osa areas, which run on to Ajah, give about .
- Two counts of verified drivers. The cohort summary and the funnel use slightly different cuts (sign-up role versus driver profile created), so their verified-driver counts differ by a handful. The rates are the same to within a point.
Caveats and what I'd do differently
- Stated intent, not trips. These are where people say they commute, captured at sign-up. Booked trips will tell the true story; they were only starting when this was run.
- Cohort, not population. The people who signed up were reached by our own marketing, so the map partly shows where we advertised. It is a demand map for Conductor, not a census of Lagos commuting.
- Grid cells, not routes. Straight lines between cell centres hide the real road choice (Third Mainland Bridge versus Eko Bridge, for example), which decides whether two people can share a car. Next step: route-level matching on the road network.
- Times of day are missing. Two people on the same corridor who leave an hour apart are not a match. I would add departure time to the intent question before reading too much into the flows.