What it really costs to sell on a marketplace

Commission is the cost everyone quotes

Marketplace profitability is decided by a stack of deductions in which the headline commission is rarely the largest item. A seller who plans on gross margin and commission alone is working from a number that will be wrong by the time the quarter closes.

Ask a marketplace seller what the platform costs them and you will usually get one figure: the category referral rate. It is the number quoted in every onboarding document and the only one most sellers can recite. It is also the one cost that is fully predictable, fully disclosed and charged only on completed sales — which makes it the least dangerous item in the stack. Everything below it is variable, partly invisible, and charged whether or not the unit sells.

Where the money actually goes

The deductions that do the damage share a property: they are not per-sale. Storage is charged on inventory sitting still, so a slow line pays rent while producing nothing. Returns cost the outbound and return shipping, the handling, and frequently the unit itself if it comes back unsellable — and on some categories the platform retains part of its fee on a returned order. Advertising has become a de facto access charge, because organic visibility on a mature marketplace is thin enough that many categories cannot be sold without it. Add currency conversion, cross-border tax handling and chargebacks, and the total deduction on a unit routinely runs to twice the referral rate a seller would name.

Build the number from the bottom up

Compute contribution per unit from the bottom up, in the order the money leaves.

Start at the selling price and subtract in sequence: referral or category commission, fulfilment, outbound shipping if separately charged, the returns provision expressed per unit sold rather than per return, storage allocated by how long the unit actually sits, advertising spend divided by units sold rather than by units clicked, payment and currency costs, and finally landed product cost including duty. What remains is contribution. It is not profit, because it carries no overhead, but it is the only figure that tells you whether selling one more unit makes you better off.

The one figure to run the business on

Run the business on contribution per unit and on the ratio of advertising spend to revenue, and look at both by product rather than in aggregate.

Aggregate numbers hide the failure mode that matters. A catalogue can look healthy overall while a third of its lines contribute nothing, because the winners cross-subsidise them invisibly. Sorting by contribution per unit, smallest first, usually produces an uncomfortable and immediately useful list. The advertising ratio deserves separate attention because it drifts upward slowly as competition increases, and a line that was viable at eight per cent of revenue is frequently not at eighteen.

How the arithmetic goes wrong

Three arithmetic errors account for most of the damage, and all three flatter the result.

The first is the returns provision: sellers deduct the cost of returns from the returned units rather than spreading it across all units sold, which understates the cost on every line with a return rate above a few per cent. The second is treating advertising as a marketing budget rather than a cost of sale — it belongs in contribution, because in most categories it is what buys the placement the sale required. The third is ignoring storage on slow lines, which converts a product with thin positive contribution into one that loses money the longer it exists. Each of these makes a weak line look adequate, which is why weak lines survive.

What this looks like on one product

In practice this is one spreadsheet with one row per SKU and about twelve columns, rebuilt from actual settlement reports rather than from the fee schedule. The settlement report is the point: it shows what was actually deducted, including the adjustments and reimbursements that never appear in a rate card. Most sellers who do this exercise for the first time find at least one line where contribution is negative and one where it is far better than they assumed, and both discoveries change what they do next.

The case for not doing this per unit

The reasonable objection is that per-unit accounting is disproportionate for a small catalogue with thin margins and limited time. If you sell forty units a day across six lines, the effort of maintaining SKU-level cost allocation may exceed the value of the insight, and a simple monthly view of revenue against total platform deductions will tell you most of what you need.

That holds while the catalogue is small and stops holding at the point where lines start behaving differently from one another — which is usually sooner than expected, because return rates and advertising costs vary far more by product than sellers anticipate. The compromise that works is to do the full calculation once, properly, on every line, and thereafter to maintain it only for the lines that move: the top sellers, and anything whose advertising ratio or return rate has changed materially since the last review.

Advertising is a commission you negotiated with yourself

On a mature marketplace, organic placement in a competitive category is scarce enough that paid placement is effectively the price of being seen. That makes advertising spend a cost of sale rather than a discretionary investment, and it belongs inside contribution alongside the referral fee.

The useful way to express it is as an effective commission: advertising spend divided by revenue, added to the platform's own rate. A category charging twelve per cent commission where you spend fifteen per cent of revenue on ads is a twenty-seven per cent channel, and that is the number to compare against your alternatives.

Expressed that way, the decision about whether a marketplace is worth selling on becomes answerable. Expressed as commission plus a marketing budget, it stays permanently ambiguous, which is why so many sellers cannot say whether a channel is working.

Returns are a per-unit cost, not a per-return cost

The instinct is to treat a return as an event that costs money when it happens. Accounting for it that way makes the cost invisible in pricing, because it lands on a different unit from the one being priced.

The correct treatment is a provision: estimate the return rate for the line, multiply by the full cost of a return — outbound shipping, return shipping, handling, any unrecoverable product value, and any platform fee not refunded — and divide across all units sold. That figure goes into the price of every unit.

The effect on categories with high return rates is large enough to change decisions. Apparel and footwear frequently carry return rates that make a line unviable at a price that looked comfortable, and sellers who price without the provision discover this only at the end of a season, when the stock has already been bought.

Read the settlement report, not the fee schedule

The fee schedule describes what should be charged. The settlement report describes what was. The two differ in ways that matter: dimensional weight reclassifications, category reassignments, storage surcharges applied in peak periods, adjustments and reimbursements that appear weeks after the order.

Sellers who build their unit economics from the published rate card are modelling a version of the platform they do not trade on. Building from settlement data is more work once and materially more accurate thereafter, and it surfaces the recoverable errors — misclassified sizes, incorrect weight bands, lost inventory — that most sellers never claim.

It also changes the negotiating position. A conversation with a platform account manager that starts from your own settlement analysis goes differently from one that starts from a complaint, and it is the only version that produces a specific correction.

Rebuild your numbers in a week

Rebuild one product's economics from settlement data before touching pricing, assortment or advertising.

Day one, export the last full quarter of settlement reports rather than working from the rate card. Day two, take your highest-volume line and subtract every deduction in the order the money leaves, arriving at contribution per unit. Day three, do the same for your slowest line, and allocate storage by the days it actually sat. Day four, express advertising as an effective commission and add it to the platform rate. Day five, sort the catalogue by contribution per unit and decide what to reprice, what to stop advertising, and what to stop stocking.

Re-run it when the ratio moves, not on a calendar

The trigger for redoing this work is not the end of a quarter; it is a change in the two inputs that drift — the advertising ratio and the return rate. Both move gradually and neither announces itself, and a line that was comfortably positive can cross into negative contribution without any single event that would prompt a review. Set a threshold for each and check monthly against it rather than reading the whole model again.

Rebuild fully when you add a marketplace, a market or a fulfilment method, because each changes the deduction stack rather than a number inside it. Cross-border selling in particular introduces currency conversion, import duty and tax handling in a way that makes an existing model inapplicable rather than merely inaccurate.

Editorial conclusion

The commission rate is the only marketplace cost that behaves well: disclosed, predictable, and charged only when you get paid. Everything else in the stack is variable, partly hidden, and frequently charged on units that never sell. Sellers who know their contribution per unit make different decisions about assortment, price and advertising from those who know their commission rate, and over a season those decisions are the whole difference between a channel that funds the business and one that consumes it.