Contents
On 2 September 2026, Uber told drivers and riders in Nigeria and Uganda that it was closing. The notice and the closure were the same day. Twelve years in Lagos, ten in Kampala, and the channel was gone by the evening.
Almost nobody reading this drives for Uber. It matters anyway, because the exposure is structural. A great many businesses in African markets get their customers through something they do not own: a marketplace, a delivery app, a social commerce storefront, a messaging platform. Those platforms are visibly narrowing. The useful question is not whether to use them, but what you still hold on the day one of them stops.
What is happening?
Uber’s statement was short: “After a thorough review, we have taken the difficult decision to wind down operations in Nigeria and Uganda, effective September 2, 2026. This decision is limited strictly to these two markets and does not impact our operations across the rest of the continent.” It said the Nigerian decision was unrelated to the Federal Airports Authority of Nigeria directive on e-hailing at airports, and that rider support would stay open for 21 days (Techpoint Africa, 2 September 2026).
That leaves Uber in four African markets — Egypt, Ghana, Kenya and South Africa. It follows withdrawals from Côte d’Ivoire in 2025 after six years and Tanzania in February 2026 after nearly ten, the latter attributed to regulatory difficulty and pressure on profitability (Channel Africa, 3 February 2026). Four markets closed in under two years.
What happened to the drivers is worth reading closely. Uber’s position was that “as independent contractors, drivers are not eligible for severance; however, as a gesture of gratitude, we have issued discretionary goodwill payouts to active drivers” (Technext). Techpoint Africa reported on 5 September 2026 that eligible drivers — those active for three to six months — saw ₦40,000 credited on 4 September, with debts owed to the platform deducted. The drivers’ union disputed the amounts. Uber did not disclose how many drivers were paid, or the criteria.
The same pattern is visible in e-commerce, with better disclosure, because Jumia files with the U.S. Securities and Exchange Commission. On 16 October 2024 it announced the closure of South Africa and Tunisia by year end, stating that “both businesses account for a negligible portion of our overall operations” and that “competitive and macroeconomic conditions in both markets have limited each country’s growth potential”. The two were 3.5% of orders and 4.5% of GMV in 2023, and 2.7% and 3.0% in the first half of 2024. Jumia then left Algeria: its first-quarter 2026 filing (7 May 2026) records that reported KPIs “exclude results from Algeria, which was exited in early 2026”, with roughly $0.8m of one-time termination benefits.
What remains uncertain: Uber publishes no country-level results, so nobody outside the company knows what Nigeria or Uganda contributed, or whether either was loss-making. The payout figures are unaudited and contested, and the market-share numbers circulating since the exit are trade press estimates we have not relied on.
Why it matters economically
A platform is a customer acquisition channel whose price and continued existence are set by someone else. That is not a criticism — it is the deal, and often a good one. But the terms can move in three directions, and you control none of them.
The rent can rise. From 1 October 2026 the WhatsApp Business Platform began charging for service replies beyond a monthly free allowance, where they had been free inside the 24-hour window; we covered what that does to the economics of messaging-led selling when it was announced. The terms can change — commission, ranking, payout timing, eligibility. Or the landlord can leave.
Jumia’s numbers show how that last decision gets made. A market worth 2.7% of orders is cheap to close, however much it matters to the sellers inside it. Your business is not a line in that calculation. It is a rounding error inside a line.
The exposure, then, is not the commission rate. It is concentration multiplied by portability. If 60% of your revenue arrives through one platform and you keep nothing outside it, a decision taken in another country can remove most of your demand with no notice — Uber’s was zero days; Jumia gave its sellers about ten weeks. If you hold the customer identity, the order history and a working direct channel, the same decision is a hard quarter rather than an extinction event.
What it means for African businesses
This is not one continental story, and the map matters.
Uber kept Egypt, Ghana, Kenya and South Africa and left Nigeria, Uganda, Tanzania and Côte d’Ivoire. Jumia left South Africa, Tunisia and Algeria while Nigeria, Ghana, Egypt and Côte d’Ivoire remain among the markets it reports on. The same country can be core to one platform and expendable to another, and a platform’s presence in your market tells you nothing about its presence in the one next door. Anyone planning expansion across the markets we work in should check platform coverage country by country rather than assuming continuity.
The exposed businesses are those whose demand arrives pre-packaged: restaurants and retailers on delivery and marketplace apps, logistics and trade operators taking jobs through a dispatch platform, sellers running on social commerce, and any business whose customer conversation lives inside a messaging app. Drivers are the clearest case because they are the purest one — a single channel, no customer list, no brand of their own, and contractor status that turned a decade of work into a discretionary payment.
Less exposed are businesses with direct contracts, named B2B accounts, physical footfall or a list they built themselves. Most firms sit in between and have never measured where.
Operating conditions sharpen the design question rather than creating it. Thin margins make commissions heavier, currency movement changes a global operator’s view of a market faster than it changes yours, and local rules on fares, commissions or licensing can make a market unviable for an operator run from abroad while leaving it viable for a business run from inside it.
Technology and AI implications
Be clear about what technology cannot do.
Software cannot recreate demand. A website does not generate the traffic a marketplace had. A CRM does not rebuild a rating history, and reputation does not port: five years of reviews sit in someone else’s database and leave with it. AI contributes nothing at the moment of loss. No model recovers a customer relationship you never recorded.
What technology does well is capture, and capture only works before the event. The practical mechanism is unglamorous: every order, booking or job that arrives through any channel writes a row into a system you control — source, date, items, value, delivery area, and the identity fields the platform and the law actually permit you to keep. That is the same discipline as treating a payment rail as a source of records rather than just money, which is why reconciliation, not the payment method, is the business system.
Two honest constraints. Many platforms deliberately mask customer contact details, so there may be nothing to capture beyond an order record. And data protection law decides the rest: consent for direct marketing must be collected lawfully at the time, not reconstructed afterwards, and regimes across the continent are tightening. A list you cannot lawfully use is not insurance.
Where AI genuinely helps is after capture: turning unstructured order confirmations and message threads into structured records, deduplicating customers who appear under three spellings, and ranking who is worth the cost of direct re-acquisition. Useful work. Not a strategy.
Practical business implications
Five questions, answerable this month.
What share of revenue arrives through each channel? Measure it monthly, by platform. Most businesses guess, and guess low.
What do you hold outside each platform? Line by line: customer identity, a lawful contact route, order history, payment history, reputation. Mark each held, partial or not held.
What is your export route, and when did you last run it? Find the export or API in each platform’s seller tools and run it now — an untested export is not a backup. Read what the agreement says about notice and about data on termination.
What is the channel costing? Commission, listing and payment fees, plus any discounting the platform requires, as a share of order value. That is your acquisition cost through it. Compare it with a direct order.
What would it take to reach your best customers directly? Not all of them — the top fifth. If the answer is “we have no way to contact them”, that is the finding worth acting on.
The aim is not to leave the platform. Platforms are efficient and most businesses should keep using them. The aim is to never be unable to leave.
RevenueStack perspective
Our view is that the recurring mistake is a category error: treating a platform as a partner when it is a landlord. Partners share downside. Landlords set the rent, and this one holds the only break clause.
Read economically, the exits of 2024–2026 are one movement: global operators consolidating into fewer African markets while raising the price of the ones they keep. Both halves push risk down the chain. Nothing about that is unique to this continent, but the concentration is sharper where a single platform can account for most of a sector’s online demand.
The design response is cheap done early and impossible done late. Build the capture layer before the replacement channel. A customer record you own, written automatically at every transaction whatever the source, costs very little and is the only thing that turns a platform exit from a cliff into a cost. The direct channel comes second: it is far easier to build a storefront for customers you can contact than to find customers for a storefront nobody knows about. That is the sequencing we argue for in how we approach a build, and why a customer system you control is usually the first thing worth owning.
AI belongs at the end of that sequence, not the start. It is leverage on data you hold. It is nothing at all on data you do not.
Conclusion
Four African markets lost Uber in under two years, and three lost Jumia. In every case the decision was made elsewhere, on numbers the affected businesses never saw, and in one case it took effect the day it was announced. No business can control that. What a business can control is the answer to a single question: if the platform that brings most of your customers closed tomorrow, what would you still have? If the answer is a list of names, a history of what they bought and a lawful way to reach them, the loss is survivable. If the answer is nothing, the platform was never a channel — it was the business.
Sources
- Uber exits Nigeria, Uganda, to focus investments elsewhere — Techpoint Africa, 2026-09-02
- Uber Exits Nigeria and Uganda, Trims African Footprint to Four Markets — Khusoko, 2026-09-02
- Uber offers Nigerian drivers ₦40,000 ($30.28) each in conditional payouts after exit — Techpoint Africa, 2026-09-05
- Uber confirms 'discretionary' goodwill payouts to 'active' Nigerian drivers — Technext, 2026-09
- Uber withdraws from Tanzania after nearly a decade — Channel Africa (SABC), 2026-02-03
- Jumia To Cease Operations in Non-Strategic Markets — Jumia Technologies AG, 2024-10-16
- Jumia Technologies AG — Form 6-K, first quarter 2026 management's discussion and analysis — U.S. Securities and Exchange Commission (EDGAR), 2026-05-07
- Jumia Technologies AG — Form 6-K, second quarter 2026 management's discussion and analysis — U.S. Securities and Exchange Commission (EDGAR), 2026-08-12