> ## Content Index > Fetch the complete content index at: https://bernardwhite.com/llms.txt > Use this file to discover other available public pages before exploring further. # Uber is leaving Africa. The problem was never the app. - URL: https://bernardwhite.com/uber-is-leaving-africa-the-problem-was-never-the-app/ - Published: 2026-09-05T00:13:55.000Z - Updated: 2026-09-05T00:13:55.000Z - Description: Uber has exited four African markets in twelve months. Swvl went from $1.5bn to $53m in six. Different mechanisms, same underlying failure - and Accra's own numbers explain why. - Author: Bernard White Danquah - Tags: Mobility, Ghana, Analysis **Uber shut down in Nigeria and Uganda on 2 September 2026, with same-day notice. Some riders were mid-journey when the app stopped working. It was the fourth African market Uber has left in twelve months, and it leaves the company operating in four countries on the continent. The technology was never the problem.** The exits came in sequence. Côte d'Ivoire in September 2025\. Tanzania in February 2026\. Then Nigeria and Uganda together, twelve years after Uber launched in Lagos and ten after Kampala. What remains is Egypt, Ghana, Kenya and South Africa. Uber's own explanation was corporate: changing priorities, part of a restructuring involving more than 3,000 job cuts and a reallocation of capital toward autonomous vehicles. Drivers received one-time goodwill payments. Nigerian users, watching trips terminate mid-ride, started calling it the Uber rapture. I run a commuter transport company in Accra, which is one of the four markets Uber still operates in. So I have read this news with more than casual interest. And I think almost every account of it has reached for the wrong explanation. ## Was this about fuel prices and commissions? Partly, and that is the story most coverage told. Nigeria removed its petrol subsidy in 2023, and the cost of keeping a car on the road rose sharply. Drivers had complained for years about commissions of 25% and higher while fares failed to track their costs. Bolt and inDrive competed hard on exactly that margin. All true. But it explains why Uber lost Nigeria to Bolt. It does not explain why the category itself has been so difficult, or why Swvl — which was not a ride-hailing company at all — collapsed just as comprehensively. Swvl was a mass-transit startup, the model closest to what I build. It listed on Nasdaq in March 2022 at $9.95 a share, valuing it at $1.5 billion, the first Middle Eastern unicorn to go public in the US. Within six months the share price had fallen roughly 95%. By November 2022 the valuation was $53 million. It laid off 32% of its staff two months after listing, then a further 50% six months later, executed a 25-to-1 reverse split to escape penny-stock delisting, and tried to unwind a $40 million acquisition it had made the year before. Uber failed on driver economics. Swvl failed on capital discipline. Different mechanisms, and the coverage treats them as unrelated. I think they are the same failure wearing two costumes. ## What actually breaks in an African city? Both companies imported a model whose core assumption is *density of coincidence*: that at any moment, enough people want to go from roughly where you are to roughly where you are going, and a matching engine can pair you. That assumption holds in San Francisco. It holds in London. It holds because those cities have high trip density, reliable road speeds, and riders whose willingness to pay absorbs the cost of an empty repositioning leg. Accra has none of those three. And the numbers are worse than most people outside the city realise. According to Ghana's National Transport Survey 2012, cited in research by the International Growth Centre, the average commute in the Greater Accra Metropolitan Area was **110 minutes, for an average distance of 3.28 kilometres**. Read that twice. If taken as a single one-way trip it implies an average speed under 2 km/h, which is slower than walking — so it almost certainly captures total daily travel across multiple legs and waiting time rather than one clean journey. The survey does not disambiguate, and I would rather flag that than quietly launder the number. But even on the most generous reading, it describes a city where getting to work consumes close to two hours of the day to cover a distance you could cycle in fifteen minutes. That is not a matching problem. No dispatch algorithm recovers those minutes, because they are not lost to poor pairing. They are lost to congestion, to waiting, and to transfers between vehicles that do not coordinate with each other. ## Why does on-demand fail specifically here? An on-demand platform needs three things at once: a rider who wants to travel now, a driver nearby who is free now, and a price that covers the driver's full cost including the unpaid kilometres spent repositioning. In a low-density, low-speed, low-fare city, those three conditions rarely coincide. When they do not, one of two things gives. Either the rider waits — and abandons — or the platform subsidises the gap. Uber subsidised the gap for twelve years in Nigeria. Then fuel prices rose, competition compressed the take rate, and the subsidy stopped making sense against a balance sheet being repointed at robotaxis. Swvl took the opposite approach and it failed faster. Rather than subsidising density, it bought it — five acquisitions in a single year, across Germany, Turkey, Spain and Argentina. Buying route networks in four countries does not produce density in any one of them. It produces four thin networks and one large integration burden, and the public markets repriced that from $1.5 billion to $53 million in about half a year. Subsidise density or acquire it. Both are attempts to manufacture a coincidence the city does not naturally supply. ## What does the city supply instead? Repetition. The same person travels from Adenta toward the centre at roughly the same time, five mornings a week, for years. They are not making a novel trip that needs solving in real time. They are making the same trip they made yesterday. Accra's commuters already know this, which is why 75% of them use tro-tro and about 13% walk — together, more than 80% of all commuting in the city. The tro-tro network is not a matching system. It is a corridor system: fixed routes, known stops, no app, no dispatch. Its weakness is not the concept. Its weakness is that departure times are unknown and capacity is unguaranteed. Which brings me to the number that reframed this problem for me. ## The finding that changed how I think about this The IGC research measures how much of Accra's formal job market a worker can actually reach by tro-tro within a given travel time: - **4%** of formal jobs within 45 minutes - **32%** within 60 minutes - **62%** within 90 minutes Fifteen minutes — from 45 to 60 — moves reachable employment eightfold. Not by making anyone faster. Purely by extending the time budget. The corollary matters more. If fifteen minutes of additional travel time multiplies job access by eight, then fifteen minutes of *unpredictable* travel time destroys it just as fast. A commuter who cannot know whether the journey takes 45 or 75 minutes must budget for 75\. They do not get the 32% band. They get something closer to the 4% band, and they make life decisions — which jobs to apply for, which to decline — inside that constraint. Accra's commuters are not primarily short of speed. They are short of *certainty*. And certainty is a scheduling property, not a routing one. ## So what should be built? If the resource the city supplies is repetition, and the resource commuters lack is certainty, then the product is a schedule, not a marketplace. Fixed corridors. Named stops. Published departure times. A seat reserved in advance, on the same vehicle, at the same time, tomorrow and the day after. The rider is buying a guarantee. The operator is buying predictable load, which is the only thing that makes the unit economics work without a subsidy. This is deliberately less clever than a matching engine. It cannot flex to arbitrary trips. Someone wanting to go somewhere unusual at an unusual hour is not served, and should call a taxi. What it does have is a demand curve that is knowable in advance, which means vehicles can be sized to it and drivers can be paid from it. A model that only works while capital covers the gap between what riders pay and what drivers need is not a business. It is a subsidy with an app in front of it. Uber ran that subsidy in Nigeria for twelve years and stopped in a single day. ## What I am not claiming That scheduled transport is obviously correct, or that I have proved it. I am building it, which makes me the least objective person to assess it, and it carries real disadvantages — it is rigid, it needs corridor density before it works at all, and it fails anyone whose life does not run to a timetable. What I am claiming is narrower: the on-demand model has now failed in African cities often enough, and expensively enough, that "better execution" has stopped being a satisfying explanation. Uber had twelve years, enormous capital and the best dispatch technology in the industry. Swvl had a billion-dollar listing. Both hit the same wall from opposite directions. When strong operators fail the same way repeatedly, the honest conclusion is usually that the model is mismatched to the terrain, not that everyone executed badly. Uber still operates in Ghana. I would not assume that is permanent. --- *Sources:* [*The Register*](https://www.theregister.com/offbeat/2026/09/03/uber-rapture-leaves-passengers-and-drivers-behind-in-nigeria-and-uganda/5294152?ref=bernardwhite.com) *and* [*Africanews*](https://www.africanews.com/2026/09/03/uber-exits-nigeria-and-uganda-as-competition-and-costs-rise/?ref=bernardwhite.com) *on Uber's exits;* [*TechCrunch*](https://techcrunch.com/2022/05/30/swvl-plans-to-lay-off-32-of-its-team-two-months-after-going-public/?ref=bernardwhite.com) *and* [*Wamda*](https://www.wamda.com/2022/11/layoffs-swvl-company-risks-delisting-nasdaq?ref=bernardwhite.com) *on Swvl; commute and job-accessibility figures from* [*Chen et al. (2017), "Unequal commutes: Job accessibility and employment in Accra"*](https://www.theigc.org/sites/default/files/2017/08/Chen-et-al-2017-Final-report.pdf?ref=bernardwhite.com)*, International Growth Centre, drawing on Ghana's National Transport Survey 2012.*