Amazon raised its take without touching commission

Amazon's total take of seller sales reached 35.1% in the year to August 2026, up 31 basis points year on year, measured on a fixed panel of several hundred Amazon sellers. Total take stacks referral commission, fulfilment, storage, inbound placement and the other Amazon-charged lines against sales. Just over a third of every pound or dollar of sales leaves before cost of goods.

Referral commission held at 13.8% of sales across both years to August 2026 on a fixed panel of several hundred Amazon sellers, unchanged year on year. The entire increase in Amazon's take came from logistics, not from the headline commission rate. Watching the commission percentage alone would have shown a flat year.

FBA fulfilment fees rose from 19.3% to 19.6% of sales in the year to August 2026 on a fixed panel of several hundred Amazon sellers. Fulfilment is the bulk of the increase in Amazon's total take. On a catalogue with thin margins, three tenths of a point of sales is the difference between a profitable SKU and a break-even one.

The FBA inbound placement fee grew 11% year on year as a share of sales in the year to August 2026 on a fixed panel of several hundred Amazon sellers. It is the fastest-rising line in the fee ledger. How you split and route shipments into the network now has a measurable effect on unit economics.

FBA storage costs sellers 1.36% of sales in February and 2.84% in November on a fixed panel of several hundred Amazon sellers, measured over the twelve months to August 2026. An annual average understates Q4 storage by more than half. Budget storage by month, not as a single yearly percentage.

Cost of goods ran at 24.3% of sales in the year to August 2026 on a fixed panel of several hundred Amazon sellers, down 38 basis points year on year. Landed product cost improved slightly. That gain was smaller than the rise in Amazon's take over the same period.

Promotional discounting fell to 2.54% of sales in the year to August 2026 on a fixed panel of several hundred Amazon sellers, down 14% year on year. Sellers competed less on price than the year before. Price cuts were not the lever used to defend volume.

The refund rate improved to 3.53% of units in the year to August 2026 on a fixed panel of several hundred Amazon sellers. Measured on units refunded against units sold. The annual figure hides a sharp seasonal swing, so read it alongside the January return wave.

Cyber Monday outsold Black Friday in both 2024 and 2025 on a fixed panel of several hundred Amazon sellers, reaching 4.17 times a normal trading day against 4.02 in 2025. The gap widened in the second year. If you plan stock and ad budget around Black Friday alone, you are underweighting the bigger day.

Unit volume falls about 60% within 24 hours of Cyber Monday on a fixed panel of several hundred Amazon sellers, in both peak seasons measured. The drop is immediate, not gradual. Inventory that has not sold by Cyber Monday evening is Q1 inventory paying Q4 storage rates.

The week before Black Friday already runs at roughly 1.8 times a normal trading day on a fixed panel of several hundred Amazon sellers, across both peak seasons measured. The peak is a ten-day event, not two days. Stock has to be receivable and live before the run-up starts, not on the Thursday.

Sellers discount less on the peak days, at 1.15% of sales on Black Friday against 1.55% to 1.80% during the run-up week, on a fixed panel of several hundred Amazon sellers. Demand does the work on the peak days. The heavier discounting sits in the run-up, where sellers are buying visibility rather than converting ready buyers.

Refunds peak in the first week of January at 6.65% of units on a fixed panel of several hundred Amazon sellers, three and a half times the 1.85% recorded during Black Friday week itself, measured across the winter of 2025 into 2026. Returns do not settle until late February. Peak-season profit is not final until the wave has passed, so hold back cash accordingly.

Aligned on Black Friday, the 2024 and 2025 daily volume curves on a fixed panel of several hundred Amazon sellers never diverge by more than 0.2 index points across fourteen days. The shape of the peak repeats. That makes it plannable: last year's daily curve is a usable forecast for this year's.

Amazon moved Prime Day from July 2025 to late June 2026 and peak-day volume on a fixed panel of several hundred Amazon sellers fell about 18%. Same sellers, same measurement, different calendar slot. Treat a Prime Day date change as a demand event, not just a diary change.

Amazon sellers on a fixed panel of several hundred Amazon sellers see four annual peaks, not one, across the two years to August 2026. Prime Day, Prime Big Deal Days in October, the Black Friday to Cyber Monday block, and the mid-December shipping deadline. Each needs its own inbound plan.

Between June and July 2025 Amazon replaced the flat coupon redemption fee with a performance fee plus a participation fee, visible in the fee ledger of a fixed panel of several hundred Amazon sellers. Any coupon cost comparison spanning mid-2025 compares two different fee schedules. Split your before and after windows at the changeover rather than reading one trend line.