We build marketplaces for operators running many sellers on one storefront — catalogue, commission rules, split fulfilment, and a payout ledger where every line reconciles to the orders behind it.
Built for the World Bank in Jordan, where one Arabic B2B marketplace carried about 1,500 businesses.
If you have fewer than twenty sellers, we will tell you on the call that you do not need this yet.
What the engagement gives you. Programme scale — sellers onboarded, demand generated, months on the ground — sits with the client work in section 13.
Three ways an operator ends up running a marketplace out of a spreadsheet. You have probably lived one of them.
A per-vendor licence that made sense at twelve sellers and does not at two hundred. The commission model is whatever the vendor decided, the payout report is whatever they export, and the answer to "can it split a payment three ways" sits on their roadmap while you wait.
A single-seller store with a multi-vendor extension bolted to it. It lists fine. It falls over at the money: refunds that do not claw back commission, a seller who ships half an order, a payout run that nobody can reconcile against the bank.
The platform handles orders and a person handles money. Every month someone exports, matches, adjusts and pays by hand — and the marketplace cannot grow past the number of rows that person can check.
All three end the same way. The catalogue looks like a marketplace and the back office is still one person with a deadline.
Nobody replatforms a marketplace because listing is hard. They replatform because the end of the month has become a two-day job with a dispute at the end of it.
The buyer is refunded, the seller's balance is not adjusted, and the difference is found weeks later by someone reading a bank statement.
A basket spanning three vendors has to split into three fulfilments, three commissions and one shipping charge that somebody has to absorb. Most platforms model the first and guess the rest.
If the answer takes more than one click, the marketplace has a trust problem rather than a software problem — and trust is the only thing keeping the sellers on it.
Hosted marketplace platforms price per vendor per month. Recruiting sellers is the whole job, and every one you win raises the bill before they have sold anything.
Category rates, introductory rates, a different split for a strategic seller — the rules that make a marketplace viable are usually the ones held back for the top tier.
Vendor accounts, payout schedules and dispute flows belong to the platform. If you leave, the relationship you spent two years building is the part that does not export.
A marketplace designed in California assumes every vendor has a company, a bank account and a card on file. Across Jordan and the Gulf a large share of the sellers worth having are individuals and home businesses, and the platform has to meet them where they are.
Around 70% of consumers in the region still prefer to pay the driver. On a single-seller store that is a reconciliation problem. On a marketplace it is three: the courier holds the cash, the commission is owed, and the seller is waiting. The model has to carry it rather than treat it as an exception.
Pickup from a producer working out of a home in a neighbourhood with no street numbering is the normal case here. Collection is resolved per city, against a courier who is actually paid.
The buyer-facing storefront being bilingual is table stakes. The part that decides whether a seller stays is whether they can read their own order list, and that console mirrors rather than translates.
A seller signs up, uploads a product and photographs it on the same device, standing in their workshop. If onboarding assumes a desktop and a scanner, the sellers you most want never finish it.
The honest objection to building rather than renting. Here is our answer, in writing.
If something we built stops working the way it was specified, we fix it. No support contract required, no hourly rate, no argument about whose fault it is.
The guarantee lapses on code modified by a third party — the trade for owning it outright.
A marketplace with no sellers is a website. Supply is the product; the software is how you keep it.
Every line below ran on Jedad. Listing products is the part every platform does; these are the parts that decide whether the marketplace still works in year two.
Most multi-vendor systems calculate the payout at the end of the month, from the orders that happen to be in the right state. We write a ledger entry the moment anything moves. The month-end figure is then a sum of entries that already exist, and every line can be opened.
Three ledger entries rather than one: what the buyer owes, what the seller is due, what the commission takes. Written together so they cannot disagree.
Cash on delivery means the money exists before it reaches you. The entry moves from due to held, against the courier, with the date it was collected.
The refund writes its own reversing entries — buyer, seller and commission — rather than adjusting the originals. Nothing is edited, so nothing is lost.
Sum the seller's entries for the period. That is the payment. When they ask why, the answer is the list of entries that made it.
A payout you cannot explain to the seller is a dispute you have not had yet.
Move the sliders. This is the same arithmetic the ledger performs, including the part most platforms get wrong — a refund clawing back its commission.
| Gross sales | 12,000.000 |
|---|---|
| Less returns | −960.000 |
| Net sales | 11,040.000 |
| Commission 12% | −1,324.800 |
| Commission returned on refunds | +115.200 |
| Seller is paid | 9,830.400 |
8,400.000 JOD of this sits with the courier until cash on delivery is banked. On a rented platform that line usually does not exist, and the payout goes out before the money arrives.
Illustrative arithmetic on figures you choose. Jordanian dinars to three decimal places, as the currency is actually quoted.
The people you meet on the call are the people who do the work.
Twenty years as a consultant and e-commerce specialist. Co-founder and COO of ShopGo, where she helped more than 1,000 MENA businesses scale online — many past $100K a month. She leads the technology practice.
Co-founder of ShopGo.me, acquired, which helped more than 1,000 MENA entrepreneurs launch online stores. He has since supported over 13,000 entrepreneurs across the region, and led the Jedad programme delivery on the ground.
Published 2026 pricing, cited underneath. Where a figure is a range we quote the bottom of theirs and the top of ours.
The fastest way to find out if anyone wants it.
Hosted, launch in days, no developer needed. Fixed commission model and limited control over checkout and payout logic. Priced per month with transaction fees on top, rising with volume.
Choose it if: you are testing whether the supply exists and expect to rebuild once you know.
Enterprise, and priced for enterprise.
Genuinely capable operator tooling and seller management. Licences run well into six figures annually before implementation, and both are sold on annual contracts with an integration partner attached.
Choose it if: you are a large retailer adding a marketplace to an existing chain, with a team to run it.
Cheap until the money moves.
A single-seller store with vendors bolted on. Fine for listings and for a dozen sellers. The failure mode is always the same and always at the ledger — partial refunds, split baskets, cash on delivery, and payouts nobody can reconcile.
Choose it if: you have under twenty sellers and simple, prepaid, single-vendor orders.
Costs more on day one. Costs nothing per seller, ever.
You own the commission rules, the payout logic and the seller relationship. No per-vendor fee, so recruiting sellers stops raising the bill. It takes months rather than days, and it only makes sense once supply is real.
Choose it if: you have sellers waiting, the money model is not standard, and you intend to be trading in five years.
If you cannot yet name fifty sellers who would list tomorrow, Sharetribe is the right answer and we will say so on the call. Supply is the product; we would rather you prove it cheaply than pay us to find out.
Riskiest part first, proven running, before the rest of the budget is committed. On a marketplace that is never the catalogue.
Commission, refunds, cash on delivery, split baskets and who absorbs the delivery fee — agreed in writing before any code. Most of the argument happens here, which is where it is cheapest.
Entries, reversals and a payout run you can execute against test data. Demonstrated to you before anything else is built.
Sign-up, approval queue, listing, and the screen that answers "why was I paid that". Ten real sellers on it before the storefront is finished.
Buyer-facing catalogue, bilingual, with search across sellers. The part every platform does, done last because it is the least risky.
One live cycle, reconciled to the bank with you in the room, before we call it launched.
Built for The World Bank in Jordan. Home businesses and small producers — the sellers who had never had a payout — listing to buyers across the Gulf. Nine product categories, each with its own attributes, each needing its own guidance before a single item was fit to sell.
From Umbrella500_Canonical_Numbers.md — figures as published in the World Bank report Jedad: Creating Market Opportunities for Refugee and Host Community Businesses in Jordan (2020), of which our co-founder Saleem Najjar is a named co-author. Third-party published, so they need no substantiation from us.
The marketplace was half the work. The other half was making 1,500 producers ready to be on it — which is why the pictures below are of design sessions rather than dashboards.
Every image above is from the programme itself. Read the full case study
We also built the SPARK Innovation Hub — 500+ organisations, launched in six weeks. Both clients are organisations, and both engagements were product builds delivered to commercial standards. More commercial work sits under NDA.
Enough that a per-vendor licence hurts, and enough that the money model is not standard. In practice that is somewhere past twenty active sellers with real order volume. Below that, a hosted platform is cheaper and faster and we will say so.
You do, including their accounts, their data and their payout history. It is your database. That is the difference between building and renting, and it is the thing that does not export when you leave a hosted platform.
Yes, and it is the reason the ledger exists. Cash collected by a courier is money that exists before it reaches you, so it is held against the courier and only released to the seller once banked. Platforms that treat cash on delivery as a payment method rather than a state get this wrong.
They open their console and see the entries that made the number — orders, commission, refunds, and what is still held. Most disputes stop there. The ones that do not have an audit trail behind them rather than a spreadsheet.
Longer than a storefront, and we will give you a range against your money model rather than a number on a page. The phases are in section thirteen and each is signed off before the next is billed, so you can stop after the ledger if the answer changes.
We have done it — 1,500 producers onboarded on Jedad, which was as much fieldwork as software. It is a separate engagement from the build and we will be explicit about which one you are buying.
Tell us who your sellers are and how the money is meant to move. Thirty minutes, no cost, and we will say so if a hosted platform is the right answer this year.
Or write directly: [email protected]