Your sourcing agent is paid to make the order bigger

A sourcing agency publishes its rates on its own website, which is already more transparency than most of the trade offers. Ten percent on orders between $1,001 and $3,000, stepping down to six percent once the order passes $20,001.

Read the slope. The rate drops as the order grows because the work does not grow with the order, and everybody in the business knows it. Placing a $60,000 order is not twenty times harder than placing a $3,000 one. What the tier table never does is flatten. At six percent the agent still earns $3,600 on that $60,000 order. Every working hour of their life is arranged around moving you up the table.

I work on the engineering side of this, in Shenzhen, in Mandarin. What follows is what buyer-side technical representation actually consists of, why hardly anybody pays for it directly and the cases where paying for it directly would be a mistake.

The slope points the wrong way

Take any decision that genuinely helps a buyer and check what it does to a percentage agent's income.

You redesign a machined aluminium housing as a moulded part and take 40 percent out of unit cost. The agent's fee falls 40 percent. Your distributor in Lagos reports slower sell-through than forecast, so you cut the order from 5,000 units to 3,000. The fee falls again. You find a second factory in Longgang quoting 18 percent under the first. Same result. You review the samples, decide the product is not ready and cancel. The agent earns nothing at all for four weeks of work that was, in every sense that matters to you, successful.

The most valuable sentence a technical representative can say is do not buy this. Under a commission structure that sentence has a price and the person saying it pays it.

There is a quieter version of the same arithmetic that the buyer never sees. In Mandarin it is 回扣, huikou. In a factory sales office the polite form is 点数, points. I have sat through quotation meetings where the number of points was worked out across the table on the assumption that the foreigner would not follow the conversation. Two points, three points, paid by the factory to the buyer's own agent, folded into a unit price that reaches the buyer's inbox looking like a single clean number.

I am not claiming this is universal and I have no figure for how common it is. The structure invites it, in the way estate agency and mortgage broking invited it in other markets, and a buyer who has never asked the question has answered it by default.

The most valuable sentence a technical representative can say is do not buy this, and under a commission structure the person saying it pays for it.

Specification before quotation

Most of the sourcing failures I have watched were decided before a single quotation came back, at the moment the buyer sent out an enquiry describing what they wanted in prose.

A specification is a controlled document with a revision number on it. Dimensions carry tolerances. Materials are called out by grade, because aluminium is not a material and 6061-T6 is. Every performance claim names the test, the standard the test is run to and who witnesses it, since waterproof is marketing and a witnessed ingress test is a fact. Packaging is a drawing with a drop height. The revision history exists so that when a factory says the drawing changed, there is a document that can say whether it did.

Then there is reading the quotation itself, which is where an engineer earns most of their keep. 含税 or 不含税, tax inclusive or not, which decides whether you are dealing with a legal supplier or only a physical one. 打样费, the sampling charge. 开模费, the tooling charge, and whether that number buys you the tool or only the parts that come off it. 起订量, the minimum order, frequently negotiable and almost never negotiated by buyers who assume it is fixed.

None of this is billable under a commission model, because nothing is billable until an order exists. The specification review is the least visible work in the sequence and the most decisive, and all of it happens before the money moves.

What a factory audit is actually looking for

验厂, yanchang, the factory audit. The certificate check is theatre. Anybody can hang an ISO 9001 certificate on a wall and most people have.

Five things are worth the trip. Whether this is the plant that will run your job or a sales office with access to somebody else's workshop, which you establish by asking for the building, the line and the shift. Whether the counterparty manufactures or trades, which is written on the 营业执照, the business licence, under 经营范围, the scope of business, in characters, for free. What happens to incoming material that arrives without a certificate. Where rejected parts physically go, because with no locked quarantine area the rejects go back into the box. And what else is scheduled on your line in the month you need parts, which in practice means asking about October, when the whole country comes back from the National Day holiday carrying a backlog.

I once walked a floor in Dongguan where the quality certificate on the wall was current and the calibration stickers on the measuring equipment had expired fourteen months earlier. Both things were true at once. Only one of them was going to show up in my parts.

Sealed samples and the first shots off the tool

封样, fengyang. A sealed sample, signed and dated across the seal, one held by each side. It sounds like ceremony until the day you need it.

The complaint that ends more supply relationships than any other is 大货和样品不一样: the production goods are not the same as the sample. Without a sealed sample in a bag with two signatures on it, that sentence is an opinion, and opinions do not win arbitration.

Tooling is where the money and the leverage sit, and the questions are dull and expensive to skip. Who owns the tool once the tooling charge is paid, in writing. Which building it physically sits in. What life it is rated for, expressed in shots rather than in adjectives, and what happens when it reaches that number. Whether the T1 trial parts get measured against the drawing or merely looked at. Whether an engineering change after T1 is billed, and at what rate.

Steel is soft before it is cut. Once a tool exists every change costs money, and once the first production run has shipped every change costs money plus time plus the argument about whose fault it was.

AQL 2.5 does not mean 2.5 percent

Here is a calculation any buyer can run for themselves. A shipment of 4,000 units, inspected at General Inspection Level II against an acceptance quality limit of 2.5, gives sample size code letter L: 200 pieces pulled and inspected, the lot accepted at 10 defects and rejected at 11.

Ten defects in 200 pieces is five percent of the sample, and the shipment passes. AQL is a property of the sampling plan, not a warranty on the goods. I have had this conversation with importers who sincerely believed they had contracted for a maximum of 2.5 percent defective units, and the misunderstanding is reasonable given how the term gets sold to them.

The tables come from ANSI/ASQ Z1.4-2008 and the ISO 2859 series. Their ancestor, MIL-STD-105, was a United States defence standard cancelled in February 1995, its final revision dated 1989. The statistical logic underneath most import contracts signed this year was settled thirty-seven years ago.

The real negotiation is the defect classification. Critical, major, minor, with critical normally set at zero. Agreeing 2.5 major and 4.0 minor takes thirty seconds. Agreeing what counts as a major defect on a moulded housing with a visible weld line takes an afternoon, needs photographs and has to be attached to the purchase order before production starts, because once 4,000 units exist you have already lost that argument.

The economics are worth stating plainly. One large third-party inspection firm publishes product inspections from $419 per man-day and supplier audits from $669 per man-day for the zone that covers China. Six percent commission on a $60,000 order is $3,600, roughly eight man-days of accredited inspection. Most buyers paying that commission are also paying for the inspection separately.

$18.9 billion against $9.57 billion

For 2024, UN COMTRADE records China exporting $18.9 billion of goods to Nigeria. The same database records Nigeria importing $9.57 billion from China. Close to half the trade goes missing between the two ledgers.

Trade economists offer several candidates and they are not exclusive: transhipment through Lomé and Cotonou, under-declaration at the port of entry, invoicing routed through third countries. Whatever the mix, the gap says something about the conditions the average Nigerian importer buys under. Money leaves in dollars, months before the goods exist. The buyer in Alaba or in Computer Village is purchasing a promise, in a currency they had to work to obtain, from a factory they will never see, under a contract no Lagos court can practically enforce in Guangdong.

Which is the corridor's real problem. The buyer with the most to lose from a bad order is the buyer least able to pay in advance for the judgment that would prevent it. No fee schedule fixes that.

I came to this as an engineer rather than a trader and it changes what you notice. A trader reads a quotation and sees a price. I read it and see which tolerances were left out so the price could be quoted at all. Living at the Shenzhen end and travelling to the Lagos end means hearing the specification discussed in Chinese and the failure described in English, and the distance between those two conversations is where most of the money goes.

The honest case for paying commission

The commission model is not a swindle surviving on buyer naivety. It solves a problem the fee model does not.

It costs nothing if nothing happens. A first-time importer cannot evaluate an engineer's judgment by reading a proposal, and knows it. Faced with an expert whose quality is unobservable before purchase, refusing to pay upfront and paying contingently instead is a rational response rather than an ignorant one. You cannot assess judgment in advance. You can assess a percentage.

Many agents also carry real risk. They take title, consolidate shipments from four factories, pay deposits out of their own account and absorb the loss when a supplier disappears. That is financing and logistics, and financing has always been priced on value rather than on hours.

They know prices, too. An agent moving 200 orders a year through thirty factories in Bao'an holds current, granular price information that somebody running twelve engineering projects a year cannot hold. Price discovery is genuine value and the fee model delivers it badly.

Then the arithmetic. Nine percent on a $6,000 order is $540. A specification review, a factory audit and a pre-shipment inspection cannot be bought for $540 anywhere on earth. Below roughly $30,000 of order value the fixed cost of doing the work properly swallows the order, and at that size the commission agent is the only option that exists.

The fee model has other dead zones. It adds little where there is no design content: buying standard 18650 cells or generic power supplies against a published datasheet calls for price discovery and supplier vetting, because the specification already exists and somebody else wrote it. It fades on repeat orders, where after tooling sign-off and three clean shipments the honest advice is to reduce scope, a sentence no percentage structure has ever contained. And it asks the buyer to make decisions. Plenty of buyers do not want to delegate execution. They want to delegate the decision, and a fee-based engineer keeps handing it back.

Prevention has no invoice

The structural difficulty with charging fees for technical work is that success is invisible. The container arrives, the goods pass, the retailer sells them and nothing happened. There is no line item for the tool that did not have to be recut, the certificate that turned out to be real, the 4,000 units that were not scrapped in Apapa. A buyer's memory of a well-run order is that it was easy, which is indistinguishable from their memory of an order that was easy because it was simple.

And the fee model tilts in its own direction, which I would rather say than have said to me. Paid by the day, there is an incentive to take more days. Paid for judgment, there is a temptation to perform judgment: a third sample round nobody needed, a tolerance pulled tighter than the application requires, a second audit booked because the first was inconclusive in a way that happened to flatter the auditor. I have caught myself on the last one.

Changing the payment structure does not remove the conflict. It relocates it, from the size of the order to the length of the engagement, and the buyer's remaining job is to know which conflict they are living inside.

What has not shifted in eight years is who pays when the answer is wrong. The agent is in Shenzhen. The container is in Apapa. The money left in March.

Sources

Leeline Sourcing, published sourcing commission schedule and service pricing : https://leelinesourcing.com/pricing/

QIMA, quality control pricing per man-day for inspections and supplier audits : https://legacy.qima.com/quality-control-pricing

QIMA, AQL calculator and acceptance quality limit reference : https://www.qima.com/aql-acceptable-quality-limit

Wikipedia, Acceptable Quality Limit (ANSI/ASQ Z1.4-2008, defect classification) : https://en.wikipedia.org/wiki/Acceptable_quality_limit

Wikipedia, MIL-STD-105 cancellation and successor standards : https://en.wikipedia.org/wiki/MIL-STD-105

Trading Economics / UN COMTRADE, China exports to Nigeria 2024 : https://tradingeconomics.com/china/exports/nigeria

Trading Economics / UN COMTRADE, Nigeria imports from China 2024 : https://tradingeconomics.com/nigeria/imports/china