Fixed costs do not care how small the order is. An entry is an entry, a broker charges a minimum, a courier or consolidator has a floor, and a test order of 200 units spreads those fixed amounts fifty times thicker than a production order of 10,000. The test unit's landed cost is therefore not a forecast of the product's landed cost. It is a measurement of the fixed cost structure, wearing the product as a disguise.
Split every landed cost into what scales with units, product price, duty as a percentage, insurance on value, and what does not: entry and brokerage minimums, document and handling fees, freight minimums, inspection flat fees. Model the test order honestly in both columns, then model the production order you would actually place, with the same columns, before anyone pronounces on the product's margin. Most test order horror stories are a fixed column photographed at small scale.
The production model needs its own honesty: freight per unit falls with consolidation, but warehousing, financing time, and obsolescence risk rise with order size, and none of those are in the simple model either. The decision deserves both corrections, not just the flattering one.
A test order buys information: demand, defect rates, return reasons, review language. Its excess unit cost is the price of that information, and it should be budgeted as a testing expense against the product's development, not buried in cost of goods where it will poison the margin report and bias the next meeting. Businesses that charge test costs to the product's ongoing margin kill their own pipeline quietly, one sensible looking spreadsheet at a time.
There are cheaper ways to buy some of the same information: samples by air for quality, a small first production run with the actual supplier and packing rather than a sample run with different economics, or a preorder campaign that tests demand at the real price. Each trades money for certainty differently. The test order is the most complete information and the worst unit cost, which is a fair trade when it is chosen on purpose.
Suppliers set minimums from their own setup costs, and the minimum is often negotiable against price, payment terms, or commitment to a schedule. A higher unit price for a smaller first order can be the right deal when it converts an unjudgeable test into a readable one and keeps total exposure low. Model the landed cost of the negotiated version, not just the discount lost: giving up two points of product price to halve the quantity at risk frequently wins on the full arithmetic.
Whatever the outcome, write down which costs in the file are test artefacts. Six months later, nobody remembers, and the test order's cost per unit will be quoted in a range review as the product's cost, by someone holding a printout, with complete confidence.
Related reading: How do peak-season surcharges change your landed-cost math?, How do you allocate freight across SKUs for accurate landed cost?.
Yes, and their entries carry the same compliance weight. A test order is still a commercial importation with classification, valuation, and origin obligations. Cutting compliance corners because the order is small creates the record that an audit later reads as the company's character.
Show three numbers: the test order's actual landed cost per unit, the modelled production landed cost with assumptions listed, and the total cash spent to learn what the test taught. The first number alone misleads, the second alone is a promise, and the third is what the information cost. Together they support a real decision.
Partly. Consolidating a test order with other cargo removes the freight minimum problem, and shipping samples by courier trades a high per kilo rate for zero fixed entry cost on low value shipments where the channel allows. Neither fixes brokerage and entry minimums on a formal entry, which is why the fixed column needs its own line in the analysis.
The smallest quantity at which the variable economics are representative and the fixed costs stop distorting the read, balanced against the demand evidence the test produced. It is a judgement, but it should be made from the split model, not from the supplier's standard MOQ accepted as a law of nature.