By Boris Dzhingarov
Most business systems assume a few suppliers and many customers. Invert that ratio and routine work like paying invoices becomes the binding constraint.
The company I run has around twenty people on the payroll and more than ten thousand suppliers on its books. The suppliers are independent website publishers who sell editorial placements to our clients. One placement can cost fifty dollars or ten thousand, though most sit between two hundred and a thousand. So the company processes a large volume of small purchases, from a very large number of counterparties, in currencies and jurisdictions we do not choose. None of that is unusual for a marketplace business. Almost all of it is unusual for the software and the habits that finance teams inherit.
The ratio the software assumes
Procurement thinking, and nearly all the tooling built around it, starts from a company with many customers and few suppliers. You negotiate annual terms with a short list of vendors. You know their account managers by name. A contract review is a scheduled event with a meeting attached. Under those conditions it is rational to spend real effort on each supplier relationship, because there are forty of them.
A whole category of business inverts that ratio. Marketplaces, freight brokers, media buying networks, creator platforms and any agency reselling third-party inventory all buy from thousands of small counterparties and sell to a far smaller number of clients. The supplier ledger is longer than the customer ledger, often by two orders of magnitude, and each supplier relationship is worth a few hundred dollars a year rather than a few hundred thousand.
Everything downstream of that inversion behaves differently. Onboarding a supplier cannot involve a call. Verification cannot involve a credit check. And the finance function, which in a customer-heavy business is a support layer, turns into the operational bottleneck, because it is the only part of the company that has to touch every single counterparty individually.
Payment is harder than sourcing
Finding suppliers was never our problem. Paying them is.
Ten thousand publishers means ten thousand preferences about how money should arrive. Some want a bank transfer, some want PayPal, some want a payment platform popular in their country and nowhere else, some want cryptocurrency. Multiply that by the transfer fees on a two hundred dollar payment, which are not trivial in percentage terms, and by the exchange spread, and by the fact that a failed transfer does not announce itself. It sits quietly until the publisher emails to ask where the money went.
Reconciliation is the part that scales worst. When a client buys a placement, the order has to be matched to a specific publisher, a specific invoice, a specific outgoing payment on a specific rail, and a specific piece of published content. Any one of those four can drift. The publisher changes the URL. The invoice arrives with a different company name on it than the one we hold. The payment goes out from a different account because the first one hit a limit. None of these is dramatic on its own. At the volume we run, a small percentage of drift becomes a full-time job.
Fraud is the same problem in a sharper form. We have paid for a placement that later turned out not to be what it appeared, and recovering that money meant a chargeback process with a payment provider who has no view of our industry and no reason to hurry. You cannot prevent this at ten thousand suppliers. You can only decide in advance how much of it you are willing to absorb, and build the process that recovers the rest.
Supplier data is a feed, not a record
The second thing that breaks is the assumption that supplier information is stable enough to store.
We sync inventory from marketplace partners through their APIs. In one of those integrations, the price shown in the catalogue matches the real price about nine times out of ten. Sites tagged as French sometimes publish in another language entirely. That feed is not badly built. Nine out of ten is a reasonable standard for a data product. It is simply not a standard you can pass through to a client without checking, because the tenth case is the one that becomes a refund conversation.
The practical consequence is that a supplier record has to be treated as a claim with a timestamp rather than a fact. Price, language, publishing policy and even ownership all move without notice, and the supplier has no obligation to tell you. So verification cannot be a one-off step at onboarding. It has to sit at the point of purchase, every time, which means it has to be cheap enough to run thousands of times a month.
Where the engineering has to go
The instinct in a business like this is to spend engineering effort on the customer-facing side, because that is what wins deals. We have put most of ours into the internal system instead: a private API that holds the supplier catalogue, ingests partner feeds, and keeps orders, invoices and payments tied to each other rather than living in three separate places.
That choice is easier to justify once you accept a specific fact about supplier-heavy businesses. Margin is not lost on price negotiation, because the individual amounts are too small for negotiation to move much. It is lost on the operational tax: the hours spent chasing a payment that failed silently, the placement bought twice because two systems disagreed, the refund issued because a catalogue was stale. Those costs grow with supplier count, not with revenue, which is why they can rise while the business looks healthy.
The useful early warning is a headcount ratio. If the number of people handling supplier admin is climbing in step with the number of suppliers, the operational model has not been built yet, and no amount of hiring will build it.
Conclusion
Leaders in supplier-heavy businesses tend to underinvest in the back office because it does not feel like strategy. It is worth checking the ratio in your own company. If your supplier ledger is longer than your customer ledger, then the systems you bought and the management habits you learned were designed for a different shape of business, and the gap between them will show up as margin you cannot account for.
Boris Dzhingarov
