Marketplaces charge different take rates by category. How to build landed cost math that survives the fee schedule.
A brand selling across five marketplace categories with take rates from six to fifteen percent does not have one marketplace cost; it has five. Blending them into an average fee produces a landed cost that is wrong for every SKU: it overstates cost in the cheap categories and understates it where fees bite hardest. Decisions made on the blend, which categories to push, where to advertise, what to discontinue, inherit the error.
The distortion compounds because the high-fee categories are often the ones with the most competition and the thinnest product margins. Apparel at fifteen percent referral plus fulfillment fees is a fundamentally different business than home goods at eight. A single blended model tells the brand these are the same business. They are not, and the P and L knows it even if the spreadsheet does not.
Start below the marketplace layer with true product landed cost: unit cost, allocated inbound freight, duty by HS code, and any category-specific compliance costs. This is the cost of having the unit ready to sell, and it is the same whether the sale happens on the brand's site or a marketplace. Get this right first; marketplace fees sit on top of it.
Then apply the marketplace fee stack in order, because order matters: referral fees typically apply to the total transaction including shipping charged, closing fees are per item, and fulfillment fees depend on dimensions and weight. Model each fee against the right base. The classic error is applying the referral percentage to the product price while the marketplace applies it to price plus shipping; on low-priced items the difference is material.
Storage fees punish slow movers by category, because dimensional weight interacts with the fee schedule differently for a sweater versus a supplement bottle. Returns processing fees hit categories with high return rates hardest, and the marketplace's return shipping deductions come out of the refund, not out of thin air. Advertising cost per category belongs in the model too; the take rate is only part of what the category costs.
Currency and tax add the final layer for cross-border marketplace sales. The marketplace may convert currency at its own rate, collect VAT or GST on the brand's behalf with its own timing, and the settlement reports do not always make the deductions legible. Reconcile settlements to the model monthly; the categories where reality diverges from the model are where the next margin project lives.
With per-category waterfalls in place, the decisions get crisp. Categories with negative contribution after the full stack are candidates for repricing, delisting, or moving to the brand's own channel where the fee stack is lighter. Categories with healthy margins can absorb more ad spend, which is often the highest-ROI move the brand is not making because the blended model hid the headroom.
Review the fee schedules quarterly, because marketplaces change them, usually upward, with modest notice. A model that was right in January can be wrong by June if the referral fee moved two points and nobody updated the waterfall. Landed cost is not a one-time calculation; it is a standing model with a refresh cadence, and the marketplace fee layer is the part that moves most.
Include it in category profitability analysis but keep it separate from unit landed cost, because ad spend is discretionary and variable while the fee stack is structural. Mixing them hides whether the category or the advertising is the problem.
Re-run the affected categories immediately and re-check pricing. A two-point fee increase on a thin-margin category can flip it negative overnight; the brands that catch it in week one reprice, and the brands that catch it at quarter-end eat it.
The waterfall discipline does, but the fee stack is replaced by payment processing, shipping, and returns costs. Running both models side by side is how you decide, honestly, which channel deserves each SKU.