
In 2024, failed deliveries made up 24 percent of Jumia’s gross cash-on-delivery orders, according to the company’s annual report to US regulators. That one rate explains more about live-selling economics in East Africa than any conversion headline. A live stream earns its money on orders that are delivered, paid for and kept, divided by every hour the team spends producing them. Count that each week and a stream becomes a decision you can price.
Key Takeaways
- The widely quoted live-commerce conversion figure, “approaching 30 percent” in McKinsey’s 2021 analysis, comes from platforms where the buy button and the payment sit inside the stream. A Kampala TikTok stream closes the sale hours later, on WhatsApp, by mobile money or at the gate.
- Cash on delivery leaks. Jumia reported that failed deliveries were 24 percent of its gross cash-on-delivery orders in 2024, and that prepaid orders deliver far more reliably.
- The Delivered-Shilling Ledger records five counts per stream (claims, confirmations, dispatches, kept deliveries, repeat buyers) and turns them into one number: contribution per host-hour.
- A host-hour includes styling before the stream, the stream itself, confirmation calls and the packing table. Labor and failed trips are heavy costs in a live seller’s week, and a viewer count hides both.
- Repeat buyers cost less to win than new ones. Jumia reported that 40 percent of its new customers from the third quarter of 2024 bought again within 90 days. A live seller can measure the same rate from her own order book.
Why does a 30 percent conversion rate mislead a Kampala seller?
The strongest case for going live more often comes from McKinsey’s 2021 study of live commerce. “Companies report conversion rates approaching 30 percent,” the authors wrote. At the top Taobao accounts they profiled, a session averaging 266,000 viewers produced roughly 75,000 orders, a conversion rate of 28 percent in November 2020. China’s live-commerce market reached an estimated $171 billion that year. McKinsey projected that live-commerce sales could account for 10 to 20 percent of all e-commerce by 2026, “if China’s experience is any guide.”
The concession belongs up front, because it is true. The format sells. A live room shortens the distance between seeing and buying, answers the sizing question in real time and shows the fabric moving under light. Even the slow markets have come around. In 2022, PingWest, citing the Financial Times, reported that TikTok had dropped plans to take its live shop into the United States and more of Europe because consumer awareness and acceptance in the UK remained low. By July 2026, the trade site ChannelX was reporting TikTok Shop figures showing UK live shopping sales up 55 percent in a year, with more than 6,000 live sessions a day. One butcher said a three-hour stream sold £8,000 of meat, stock that would usually take three to four days to move through a high street shop.
Every one of those numbers rests on a feature the Kampala seller is missing. McKinsey defines live commerce as the combination of “instant purchasing of a featured product and audience participation.” In China and the UK the tap is the payment. In Kampala the tap is a comment. The venue is narrower too: the LONDA 2025 digital rights report records that Facebook remained blocked in Uganda as of December 2025. Kenyan sellers face the same gap between interest and money. Business Daily reported that TikTok transactions in Kenya are often conducted informally, which leads to disputes over product quality, delivery times and payment. The conversion that pays the rent happens later, at a gate in Ntinda or a boda stage in Kisaasi, when cash or a mobile-money confirmation changes hands.
Where does a live order leak between “mine” and money?
Follow one order from comment to repeat. A viewer types “mine” under a dress. An assistant moves her to WhatsApp to confirm size, color, location and delivery fee. A packer bags it and a rider carries it across town. Then the order either lands or leaks. The buyer is home or out. She pays or she has changed her mind. She keeps it, or she sends it back because the color looked different under the ring light. Weeks later she buys again, or she drifts to the next stream in her feed.
Each step leaks, and a useful public measure of the leak comes from a marketplace with far more logistics than a six-person shop. Jumia, which operates in nine African countries, told the US Securities and Exchange Commission that failed deliveries were 24 percent of its gross cash-on-delivery orders in 2024. Its wider rate for cancellations, failed deliveries and returns was 22 percent of gross merchandise value and 23 percent of orders that year. The company wrote that prepaid orders drive “much higher delivery success rates than payment on delivery.”
Two cautions about borrowing that figure. It is a marketplace average across many countries. A live seller’s buyers have also watched her for an hour, which may make them more committed or more impulsive than a marketplace shopper. Public research on Kampala live-seller funnels is thin, and AVODA holds no cohort dataset of its own on this. Treat 24 percent as a warning light, then replace it with your own rate within a month. The broader channel picture sits in WhatsApp and TikTok commerce as Africa’s real stack; this piece stays with the arithmetic of one stream.
The Delivered-Shilling Ledger
The Delivered-Shilling Ledger is five counts a seller writes down after every stream, plus one number she computes from them on Friday.
- Claims: the “mine” comments and direct messages that name a specific item.
- Confirmations: claims that agreed size, price, delivery fee and location on WhatsApp, or paid a mobile-money deposit.
- Dispatches: parcels that left the shop with a rider.
- Kept deliveries: parcels paid for and still with the buyer seven days later.
- Repeat buyers: kept-delivery customers who order again within 90 days.
Each ratio points at a different person in the shop. Claims to confirmations measures the follow-up on WhatsApp, which the conversational commerce stack can speed up. Confirmations to dispatches measures stock accuracy: the dress was promised twice, or the size was gone. Dispatches to kept deliveries measures payment terms and honest showing. Kept deliveries to repeat buyers measures whether the stream is building a customer base or renting attention one evening at a time.
The one number is contribution per host-hour. Take the gross margin on kept deliveries. Subtract every delivery cost, including the trips that came back, plus data, airtime, giveaways and discounts. Divide by every hour the team spent on the stream. That denominator carries the styling hour before going live, the confirmation calls afterward and the packing table. A two-hour stream consumes far more than two hours of the team’s time, and the ledger counts all of them. It also gives the shop a single figure to sit beside the others on a five-number dashboard.
What does one stream actually cost?
Labor comes first. A shop that goes live five evenings a week is paying a host, someone reading comments, a stock runner and a packer, whether those roles sit on a payroll or on a family arrangement. Price each hour at what you would pay to replace that person, because the stream is spending it.
Data comes second. Uganda taxes internet data under the Excise Duty (Amendment) Act 2021, a 12 percent levy that the LONDA report says fell hardest on women’s internet access. Check how much data one stream uses on the phone’s data counter. Then price it against the bundle you actually buy, using the price comparison the Uganda Communications Commission publishes.
Failed trips come third, and they hide well. For illustration only, with round numbers: suppose a stream produces 40 confirmed cash-on-delivery orders. At Jumia’s 2024 failure rate, about 10 of those parcels ride out and ride back. The shop pays the rider both ways, the packer re-shelves the item, and the dress spends a day off the rail while another buyer might have taken it. That cost lives in the rider’s notebook, far from the stream’s dashboard.
There is a second revenue line for sellers whose audience grows large. TikTok for Business entered Kenya in 2025. It said that in its first year there, more than 200 Kenyan creators earned over Sh47 million, about $350,000, through brand collaborations, and Kilimall generated more than 152,000 purchases through TikTok advertising. TNX Africa reported those as platform-side figures, and they cover creators and brands. They describe a market forming, and they say little about the margin on a single dress.
Which levers move contribution per host-hour?
Payment terms move it fastest. Jumia’s filing states plainly that prepaid orders deliver more reliably. For a live seller, that means a mobile-money deposit before the parcel leaves, for first-time buyers or for orders above a set value. It will cost some claims. Run it on alternate streams for two weeks and compare kept deliveries per host-hour across the two sets. The ledger will tell you how many of the lost claims were ever going to pay.
Showing accuracy moves the fourth line. A 2025 study of 340 live-stream shoppers found that trust in the seller had a stronger effect on purchase intention than trust in the product. For a shop, that means daylight or true-color lighting, the label shown on camera, and any flaw named before a viewer finds it. Whether that lowers your returns is an empirical question, and line four of the ledger answers it.
Schedule is a variable to test. Seller-side research in the Journal of Marketing Management found that sellers using the same live medium experience different levels of success. It identified four sales approaches and twelve strategies for winning and keeping customers. Hold the weekday and hour steady for four weeks, then change one thing and watch the ledger.
Repeat buyers cost less to serve, because a returning buyer already knows your sizing, your rider and your word. Jumia’s 40 percent repurchase rate within 90 days is a useful outside reference. Message past buyers on WhatsApp before each stream and record how many of the night’s confirmations came from them. The case for building the business on that number is set out in retention as the cheap growth engine.
Category sets the ceiling. In McKinsey’s analysis of China, apparel and fashion took a 36 percent share of live-commerce categories, with beauty and food at roughly 7 percent each. Clothing shows well on camera, and it also comes back when it does not fit. Choose items where the camera settles the buyer’s doubt before she types it.
Honest showing and the kept promise
The ledger is an honest secular instrument, and a good one. It treats a “mine” as an intention and counts only the shillings that arrive, which is sound practice for any shop. Its limit sits inside its own findings. The order fails at the door, or the dress comes back because the color lied under the ring light, and the standard playbook answers with pressure: McKinsey notes that “time-limited tactics such as one-off coupons can be used to generate a sense of urgency.” Urgency raises claims. On the ledger’s own arithmetic, it can also raise the failed trips that follow. Jesus’ instruction, “let your ‘Yes’ be ‘Yes'” (Matthew 5:37), runs in both directions across the screen. Femi Adeleye, in Preachers of a Different Gospel, counts among the gospel’s demands “that market women no longer hide rotten tomatoes under fresh ones.” On a live stream the rotten tomato is the camera angle that hides the stain. A seller whose yes is reliable trains buyers whose yes is reliable, and the 2025 trust study points the same way: buyers weigh the seller’s word above the product. Seen that way, a deposit is a covenant with two signatures. That gives the ledger what it was reaching for and could only count: a stream where the comment and the cash are the same promise.
Frequently Asked Questions
What conversion rate should a Ugandan live seller expect?
Public benchmarks for Ugandan streams are thin. The 30 percent figure comes from McKinsey’s 2021 analysis of platforms with in-stream checkout, mostly in China, where the tap is the payment. Measure your own rate from claims to kept deliveries over four weeks of streams, and use that as your baseline.
Should a live seller drop cash on delivery?
Test before deciding. Jumia reported that 24 percent of its cash-on-delivery orders failed in 2024 and that prepaid orders deliver more reliably. Ask first-time buyers, or buyers of higher-value items, for a mobile-money deposit on alternate streams, then compare kept deliveries per host-hour over two weeks.
What is contribution per host-hour?
It is gross margin on kept deliveries, minus all delivery costs including failed trips, data, airtime, giveaways and discounts, divided by every team hour the stream consumed. That includes styling before going live, confirmation calls afterward and packing. It is the single number the Delivered-Shilling Ledger produces each week.
How often should a small shop go live?
Treat frequency as a test. Research on live sellers in the Journal of Marketing Management found that sellers on the same medium achieve different levels of success through different approaches. Hold one schedule steady for four weeks, record the ledger after every stream, then change one variable and compare contribution per host-hour.
Can live selling pay for a six-person shop?
It can, when the stream produces enough kept deliveries per team hour to cover labor, riders and data with margin left over. A stream can look busy and still lose money on failed trips. Five weeks of ledger records will show which evenings pay and which only entertain.
Next Step
Run your next three streams with the five counts on a single sheet, and compute contribution per host-hour by Friday. Then take the AVODA Business Capability Assessment to see how your margins, cash cycle and team capacity compare with the rest of the business. Doing it with this week’s numbers in hand saves you weeks of going live on a guess.
Related Reading
- WhatsApp & TikTok Commerce: Africa’s Real E-Commerce Stack
- From Chat to Checkout: East Africa’s Commerce Stack
- Retention Is the Cheap Growth Engine You’re Ignoring
- The Five-Number Dashboard for a Small Firm
Sources and Evidence
- McKinsey Digital, live commerce report on how livestream selling is changing shopping (July 2021): conversion rates approaching 30 percent; Taobao top-session figures (266,000 viewers, 75,000 orders, 28 percent); China’s $171 billion market in 2020; 10 to 20 percent projection for 2026; definition of live commerce; urgency tactics; category shares.
- Jumia Technologies, Annual Report for Fiscal Year 2024 (Form 20-F): failed deliveries 24 percent of gross cash-on-delivery orders in 2024; CFDR 22 percent of GMV and 23 percent of orders; prepaid orders deliver more reliably.
- Jumia, “Jumia Files Annual Report 2024 on Form 20-F”: Jumia operates in nine African countries.
- Jumia Reports Fourth Quarter 2024 Results: 40 percent of new customers from Q3 2024 made a second purchase within 90 days.
- ChannelX, “TikTok Shop: 300,000 UK SMEs & LIVE Shopping up 55% year-on-year” (July 2026): UK live sales up 55 percent; more than 6,000 live sessions a day; butcher’s £8,000 three-hour stream.
- PingWest, “TikTok aims to change its lukewarm live shopping business in the UK”: FT-reported retreat from US and European live-shop expansion; low UK consumer awareness and acceptance.
- Paradigm Initiative, LONDA 2025 Digital Rights and Inclusion Report: Uganda: Facebook blocked in Uganda as of December 2025; 12 percent tax under the Excise Duty (Amendment) Act 2021 and its effect on women’s access.
- Business Daily, “How TikTok is redefining commerce for Kenyans”: TikTok live selling in Kenya; informal transactions and disputes over quality, delivery and payment.
- TNX Africa, “TikTok for Business marks one year in Kenya”: 200+ Kenyan creators earned over Sh47 million ($350,000); Kilimall 152,000+ purchases; Kenya entry in 2025.
- Uganda Communications Commission, “Comparison of Mobile and Fixed Internet Prices, as of June 2025”: the regulator’s published bundle price schedule for pricing a stream’s data.
- ScienceDirect, study of parasocial relationship, trust and purchase intention in live-stream shopping (2025): 340 respondents; trust in the seller has a stronger effect than trust in the product.
- Journal of Marketing Management, “Live streaming commerce from the sellers’ perspective” (2020): sellers achieve different levels of success; four sales approaches and twelve strategies.
- Matthew 5:37, New King James Version: the wording “let your ‘Yes’ be ‘Yes'” quoted in the faith paragraph.
