Ask a seller what a unit costs them and you will usually get cost of goods plus referral fee plus FBA. Three layers out of eleven.
Here is the full stack, in the order they hit you.
1. Cost of goods. Landed, not ex-factory. If you import, the number on the supplier's quote is not your cost.
2. Referral fee. Category-banded, and the band is not always the one you assume. Mis-categorised SKUs pay the wrong rate for years.
3. Closing fee. Small, fixed, and applied per unit — which makes it brutal on low-price SKUs.
4. FBA pick and pack. Size-tier based. A packaging change of a few millimetres can move a SKU between tiers in both directions.
5. Weight handling. Often the largest single fee on heavy goods, and the one most sensitive to packaging nobody has re-measured since launch.
6. Storage. Charged by volume and by time. Slow SKUs pay this repeatedly, which is why ageing inventory is more expensive than its carrying cost suggests.
7. Advertising, apportioned. Not campaign ACOS — the ad cost actually carried by that SKU, including the spend that generated a sale of something else.
8. Returns and RTO. The single largest variable in fashion and a material one almost everywhere. Use the SKU's own rate, not a catalogue average.
9. Reimbursement gap. Inventory lost or damaged that was never claimed. It is a real cost precisely because it was recoverable and was not recovered.
10. Payment gateway. Where applicable by channel. Small per unit, visible across a year.
11. GST on fees. Charged on the fees themselves. Frequently omitted, always real.
Model all eleven and two things usually happen. First, some SKUs you thought were your best turn out to be funded by the ones you were about to discontinue. Second, your break-even ACOS stops being a round number somebody chose.
The Amazon Profitability AI tool runs all eleven per SKU, in bulk, against live fee tables.