Contents
Most retailers who deliver treat cash on delivery as a concession to the customer: offer it and the order comes in, refuse it and the customer buys elsewhere. It is rarely costed. The order is counted when it is placed, the sale when it is dispatched, and the parcel that comes back three days later is written off as bad luck rather than as a line in the margin.
There is now a public number for how often that happens. Jumia, which operated in nine African markets at the end of 2025, disclosed in its annual report that failed deliveries represented 25% of the gross orders paid by cash on delivery. This piece is for the operator of a retail or e-commerce business who takes payment at the door and suspects — correctly — that the returns are eating the margin, but cannot see by how much.
What is happening?
Jumia Technologies AG filed its annual report for the year ended 31 December 2025 with the US Securities and Exchange Commission on 24 February 2026. Three of its disclosures are useful to any retailer, marketplace seller or not.
- The failure rate. “In 2025, failed deliveries represented 25% of the gross orders in Cash on delivery.” The company attributes this to customers not being present for a door delivery, or not collecting from a pick-up station, and says orders paid electronically in advance “tend to drive much higher delivery success rates than payment on delivery”.
- The combined leakage. Its cancellations, failed deliveries and returns rate rose from 22% of gross merchandise value in 2024 to 26% in 2025, and from 23% of orders to 24%. The value measure deteriorated faster than the count, which means the orders falling away were, on average, the larger ones.
- Who carries the loss. It makes three delivery attempts, then returns the goods to the seller, and must notify that seller within 21 days of shipment; if it does not, it takes possession and accepts the loss itself.
The same filing says most of its transactions are completed by cash on delivery, which it attributes to customers who either have no bank account or do not trust online payment, and that use of its own payment gateways rose only from 27.1% of order value in 2024 to 28.4% in 2025. The shift to prepaid is real, and slow.
What this does not establish is an industry-wide failure rate. It is one operator’s book, with its own category mix, delivery network and customers, and the combined measure blends cancellations before dispatch with failures at the door. Treat 25% as a well-documented reference point, not as your number.
Why it matters economically
A failed delivery is not a lost sale. A lost sale costs the opportunity; a failed delivery has been picked, packed, dispatched, carried out and carried back, tying up stock and working capital for the round trip. The cost is incurred in full and the revenue is zero.
The scale becomes visible per order. Our calculation, from the 2025 annual report: gross merchandise value of US$818.6m across 22.6m physical-goods orders is roughly US$36 an order; gross profit of US$101.8m is about US$4.50 an order, or 12.4% of order value; fulfilment expense of US$45.5m is about US$2.00 an order. In the second quarter of 2026, gross profit was 14.2% of order value, up from 13.3% a year earlier.
Those averages belong to a marketplace, not a shop — they blend own-stock sales, commissions, delivery charges and advertising — so do not import them into your accounts. The relationship between them transfers:
- The order value that cancels, fails or comes back (26%) is roughly twice the gross margin earned on everything sold (12.4%).
- The fulfilment cost of one order (about US$2.00) is close to half the gross profit of a successful one (US$4.50) — and a failed door delivery usually spends it twice, outbound and inbound.
One failed delivery therefore does not cost a fraction of a sale. It can cost the whole margin on a successful one, and at a 25% failure rate one cash-on-delivery order in four is carried twice for nothing. That is the arithmetic behind the feeling that volume is growing while the bank balance is not — before counting shrinkage on cash collected by riders and the cost of the working capital tied up in stock in motion.
What it means for African businesses
Cash on delivery persists for structural reasons, not cultural ones. Jumia’s filing names them: customers without bank accounts, and customers who could pay online and do not trust it. Both are design constraints to build around, not habits to wait out.
The alternative is also less available than account numbers suggest. The GSMA reported on 24 March 2026 that mobile money passed US$2 trillion in global transaction value in 2025 across 2.3 billion registered accounts, but that monthly activity reached only 25.7% — its highest since 2021, still leaving roughly three-quarters of registered accounts unused in a given month. That rate is global, not African, though the continent accounts for most of the growth. The reading for a retailer is narrow: having a wallet is not the same as paying with it.
The picture differs by market, so one policy across borders will be wrong somewhere. Returns can be a statutory right rather than a goodwill gesture: Jumia notes that in Egypt, one of its largest markets, customers may return any product within fourteen days of delivery without giving a reason. Where card and bank payment are already normal, prepaid conversion is a pricing decision; where they are not, removing cash on delivery removes customers. Last-mile cost also varies inside a country — the filing notes that delivery to a secondary city or rural area costs more than to a main city, and home delivery more than a pick-up station. Jumia ran 2,422 pick-up stations at the end of 2025 for that reason.
This is a problem for businesses delivering physical goods to consumers, and a different one for a logistics or distribution business, where a failed delivery costs carriage rather than margin on the goods.
Technology and AI implications
The decisive change is measurement. Most retailers cannot answer “what is my first-attempt delivery success rate, by payment method, by area?” because the order, the dispatch and the payment live in three places. Joining them is a systems job: one order record carrying the payment method, the delivery attempts, the collection outcome and the cost of each leg. Without it, every decision here is opinion.
Once it exists, a commerce system can act where the cost is created — at checkout and at dispatch. Confirmation before dispatch, a delivery window the customer picks, a deposit on higher-value orders, a pick-up point as the default for a difficult address, a discount for paying in advance: each is a configuration decision testable against the failure rate. Matching the money to the order once it does arrive is separate work, covered in mobile money is a payment method.
Where technology does not help, plainly: software cannot make a customer answer the door, create trust in online payment, or recover cash a rider did not collect. It can flag an address that was never collectable, not fix it. The largest lever is commercial — what you offer, to whom, at what price.
AI has a narrow, genuine role: normalising messy addresses, flagging orders statistically likely to fail so someone calls before dispatch, and clustering failure causes from delivery notes. It also carries a real risk. A model that declines or down-prioritises cash-on-delivery orders will be wrong about individual good customers, and where cash is the trusted default that is an expensive kind of wrong. Keep a person on any decision that turns a customer away.
Practical business implications
- Separate the three failures. Cancellation before dispatch, failure at the door and return after delivery have different causes and costs. One “returns” number hides which is growing.
- Cost one failed delivery properly, once. Carriage out and back, handling, packaging, days of stock tied up, and the margin that never arrived. Multiply by last month’s failures: that is the budget for fixing it.
- Measure first-attempt success by payment method and area. If prepaid orders succeed materially more often than cash-on-delivery ones — as they do on Jumia’s platform — that gap is the price of the concession, and you can decide whether to keep paying it.
- Change the mix before removing the option. A deposit above a threshold, prepaid-only for fragile or high-value categories, pick-up collection for areas with a poor record, a call before dispatch. Each narrows the exposure without closing the channel.
- Fix contactability first. A number that answers and an address that resolves prevent more failures than any pricing change.
Below roughly twenty deliveries a day, a call-before-dispatch routine and a notebook are honest. Above it, the failures are a system problem.
RevenueStack perspective
Our view: cash on delivery is not a payment method. It is a short-term credit and trust product that most retailers give away at no price, paid for in failed deliveries rather than in interest. The retailer finances the order, carries the inventory risk, pays the carriage both ways and recovers nothing when nobody is home.
Read through strategic economics, the leverage is not in transaction fees — where most attention goes — but in the mix and the failure rate. A business that moves ten percentage points of orders to prepaid, or lifts first-attempt success by ten points, changes its gross margin more than any fee negotiation will. Read through technology, the enabling step is one order record that knows the payment method and the delivery outcome, because a cost that is not attributed cannot be managed; our approach starts there, not with a platform choice. Read through AI, the honest answer is a small supervised role at the edges and none at the centre.
The detail worth holding on to is the 21-day notification rule in Jumia’s seller terms. In a marketplace, the loss from a failed delivery is allocated by contract to somebody. In your own shop there is no clause and no counterparty: it falls to you, by default, every time.
Conclusion
The question is not whether to offer cash on delivery. In most of these markets it is still how customers buy, and refusing it refuses revenue. The question is what it costs, order by order and area by area, and whether that cost is visible in time to act on. A quarter of cash-on-delivery orders failing is a documented figure for one operator across nine markets; what matters is yours, and most retailers do not know it. Cost a single failed delivery honestly this month, then let the number decide where the concession stays and where it needs a deposit, a collection point, or a call before the rider leaves.
Sources
- Jumia Technologies AG — Annual Report on Form 20-F for the year ended 31 December 2025 — US Securities and Exchange Commission, 2026-02-24
- Jumia Reports Second Quarter 2026 Results and Announces Capital Raise — US Securities and Exchange Commission, 2026-08-12
- Mobile Money accounted for $2 trillion in transactions in 2025, doubling since 2021 as active accounts continue to grow — GSMA, 2026-03-24