The Amazon Fee Creep Report (2026)

What 'fee creep' means

Amazon's take isn't one number. It's a stack: a percentage referral fee, a largely fixed per-unit fulfillment fee, storage, and (for most sellers) advertising on top. Some of those costs scale with price and some don't, and that mix is what determines how much of each sale you actually keep.

The fixed parts are the problem. A fulfillment fee of a few dollars barely dents a $100 product but devours a $10 one. That's fee creep: as your average selling price falls, the proportion of revenue lost to fixed costs climbs fast.

It helps to sort the stack into two buckets. Proportional costs (the referral fee, ad spend measured as a percentage of sales) take the same share of revenue at every price point. In this model they sum to roughly 27% of the sale whether the item costs $10 or $100. Fixed costs (fulfillment, and to a lesser degree storage) are priced on the physical box, not the sticker, so their share of revenue is entirely a function of your price. Fee creep is just the fixed bucket expanding as price shrinks. Nothing about it is a surprise once you see the two buckets separately, but almost nobody prices a product that way on the first pass.

The model: what sellers keep by price band

Applying the assumptions above, here's the modeled share of each sale a seller keeps after Amazon fees and advertising, but before the cost of the product itself.

Read the table down the Kept % column and notice where the gains come from. Moving from $10 to $20 lifts kept-share from 37% to 52%, a 15-point jump for a single price-band step. Every step after that is smaller: 5 points from $20 to $30, then 4, then 5. In other words, the returns to raising price are heavily front-loaded. Escaping the bottom band matters more than anything you do further up the curve, because that's where the fixed fulfillment fee is doing the most damage.

Also notice what happens in dollar terms rather than percentages. The modeled seller keeps $3.70 on a $10 sale and $10.45 on a $20 sale. Doubling the price nearly triples the pre-COGS dollars kept, because the fixed fees were already paid once and the extra revenue flows through at close to the proportional-fee rate. That asymmetry is the entire economic argument for bundling and multipacks.

Why cheap products get crushed

Look at the fulfillment column: it barely moves from $10 to $100, because it's priced on the box, not the sticker. That near-fixed fee is ~35% of a $10 sale but under 7% of a $100 sale. The referral fee is a flat percentage, so it's neutral across price, but the fixed fulfillment fee is what turns low-ASP selling into a margin trap.

This is why so many sub-$15 products that look profitable on a spreadsheet lose money in reality: the seller used a percentage in their head and forgot the fixed dollar fee that doesn't scale down.

Put a landed cost against the model and the trap snaps shut. The $10 seller keeps $3.70 before paying for the product, so any item with a landed cost above $3.70 is underwater in this model even before returns, damaged units, or a single remeasured dimension. A typical sourcing rule of thumb that works fine at $30 or $50 simply has no room to operate at $10. And the failure is quiet: the product sells, reviews come in, velocity looks healthy, and the bank account still shrinks, because every individual sale is a small loss dressed up as progress.

Advertising is the silent compounder

Most products don't sell without ads, and ad spend scales with price as a percentage, so it stacks on top of the fixed fees rather than replacing them. In the model, advertising alone is the difference between an uncomfortable margin and no margin on the cheapest items.

The takeaway isn't 'stop advertising'. It's that ad spend has to be judged against the margin that's *left* after fixed fees, not against the headline price. A 12% ad cost on a product that already loses 50% to fees is very different from 12% on one that keeps 65%.

The model makes that concrete. On the $10 item, Amazon's fees alone (referral, fulfillment, storage) come to $5.10, leaving $4.90 of headroom. The $1.20 of ad spend is 12% of the sale price but roughly a quarter of that remaining headroom. On the $100 item, the same 12% ad rate takes a far smaller bite of what's left after fees. Same TACOS, radically different real cost. This is why judging campaigns by ACOS or TACOS alone, without reference to the post-fee margin of the specific SKU being advertised, routinely funds campaigns that profitably grow revenue while unprofitably shrinking the business.

Why averages hide the truth, and what to do

Because kept-share swings about 30 points across these price bands, any single 'Amazon takes about a third' rule of thumb is dangerously wrong at the edges. The only reliable way to know your real margin is to compute it per SKU, with the actual fees Amazon charged on your settlements and your true landed cost.

Practically: track fee, ad, and product cost per unit; raise price or bundle low-ASP products toward the band where the math works; and audit your settlements for the fee and reimbursement errors that quietly widen the gap. That per-SKU truth is exactly what accounting built for marketplaces is for.

Pair two reports when you do that audit. The settlement report tells you what Amazon charged; the inventory ledger report in Seller Central tells you what physically happened to the units, receipt by receipt and adjustment by adjustment. A fee line that doesn't correspond to an event in the ledger is the shape most fee errors take, and you can only see the mismatch by reading the two side by side.

This is where the report connects to bookkeeping, because the blended average isn't just an analyst's mistake, it's the default output of naive books. A seller who records net deposits as revenue literally cannot see kept-share by SKU; the fees that drive the whole effect are invisible. Proper amazon fba accounting splits every settlement into gross sales and each fee type, which is the raw material for the per-SKU margin math this report argues for. Good amazon seller accounting software should be able to answer 'what share of revenue did fees take on this SKU last quarter?' without a spreadsheet export. Whether that means QuickBooks for Amazon sellers with a settlement connector in front of it, or Xero with the same, the requirement doesn't change: the ledger has to see gross sales and fee lines, not a net deposit. If your current setup can't, that's the gap to close before optimizing anything else, and pairing it with amazon reimbursement software closes the second gap: fee overcharges and lost inventory that pad the fee column without your knowledge.

• •Compute margin per SKU from real settlement fees, not category averages.
• •Push low-ASP products up a price band or bundle them to dilute fixed fees.
• •Judge ad spend against post-fee margin, not the sticker price.
• •Audit settlements for fee/dimension errors and unclaimed FBA reimbursements.

What the model implies for pricing and sourcing decisions

The practical use of this model isn't the specific percentages, which will drift as Amazon updates its fee schedule. It's the shape of the curve, and the shape supports three decisions.

First, sourcing filters should be price-band aware. A flat 'target 30% margin' screen will pass low-ASP products that can't structurally deliver it and may reject higher-ASP products that comfortably can. Screen cheap products against the punishing fixed-fee share of their band, not against a portfolio average.

Second, the model favors moving up-market over squeezing suppliers at the bottom. Shaving cents off a landed cost does nothing to the $3.50 fulfillment fee that dominates a $10 sale, while repositioning the same product as a two-pack or a bundled kit re-prices it into a band where fixed fees dilute. The front-loaded gains in the kept-share column (15 points for the first band step, roughly 5 for each later one) say the effort is worth the most at the bottom of the curve.

Third, treat the model as a hypothesis to test against your own settlements, not a conclusion. Your category's referral rate, your real TACOS, and your measured size tier all shift the curve. The structure (proportional costs flat, fixed costs inflating as price falls) will hold; the exact crossover points are yours to find with your own numbers.

Does the same math apply off Amazon?

The mechanism isn't Amazon-specific. Any charge billed in dollars rather than in percent behaves this way, so the useful question for a multichannel seller is which channels carry fixed fees at all.

eBay is the clearest contrast. You ship the order yourself, so no fixed fulfillment fee sits inside the platform's take; the final value fee is a percentage, and the eBay ad fee on Promoted Listings is a percentage as well. In the two-bucket framing from earlier, that stack is almost entirely proportional, so kept-share barely moves across price bands. This is the honest version of Amazon vs eBay for sellers at the low end: eBay's percentages may or may not be kinder, but its structure doesn't punish cheap products the way a fixed fulfillment fee does. The fixed cost hasn't vanished, though. It moved onto your shipping label, your packaging, and your hours, none of which shrink when the price does.

eBay's optional listing charges are the exception that proves the rule. Ask what an eBay reserve price fee is and the answer is a charge for setting a minimum you're willing to accept on an auction: a fee for listing, not for selling. An auction reserve price fee, an insertion fee, a listing upgrade, all of them are fixed dollar amounts, which makes them behave exactly like the fulfillment column in our table. Trivial against an expensive item, meaningful against a cheap one. If you run reserve auctions on low-ASP inventory, price those listing fees the way this report prices fulfillment.

Two more variations worth naming. Amazon FBA vs dropshipping changes which line you pay, not whether you pay it: a dropshipper never hands Amazon a fixed fulfillment fee, because the supplier ships, but the referral percentage is unchanged and the shipping cost reappears inside the supplier's price. Fee creep doesn't disappear, it just gets billed by someone else. And the Vendor Central vs Seller Central split removes the stack from your books entirely, since a first-party vendor sells wholesale to Amazon at a negotiated price rather than paying referral and fulfillment fees. Amazon has priced all of this in on its own side of that deal.

Limitations of the model

Honesty about what this model doesn't capture matters as much as the result. It excludes product cost, so it says nothing about absolute profitability, only about the share of revenue available to cover cost and profit. It assumes a constant ad percentage, while in reality cheap products often need a higher TACOS to rank, which would make the bottom bands look even worse than shown. It models standard-size products only; oversize fee tiers would change the fulfillment column materially. And it ignores returns, refund administration, and seasonal storage surcharges, all of which fall more heavily, as a share of revenue, on exactly the low-priced items the model already flags. It leaves out per-unit charges that sit outside the sale itself, too: an Amazon FBA disposal fee or removal fee on units that never sell, and whatever a prep center bills per unit to label and pack inventory before it ships in. Each of those is a fixed dollar amount per unit, so each one lands hardest on the same cheap products.

Each of those simplifications was chosen to keep the model reproducible from published fee structures alone. Every omitted factor we're aware of would deepen the low-price penalty rather than soften it, so treat the modeled kept-shares for cheap products as an optimistic ceiling, not a floor.

FAQ

How much does Amazon take per sale?

It depends heavily on price. In this model a seller keeps about 37% of a $10 sale but about 66% of a $100 sale after fees and ads (before product cost), because the fixed fulfillment fee consumes a much larger share of cheap items. Exact fees vary by category and size, so verify against Amazon's current schedule.

Why do cheap products have worse margins on Amazon?

Because the FBA fulfillment fee is largely fixed per unit. A few dollars is a small slice of a $100 item but a huge slice of a $10 one, so low-priced products lose a disproportionate share of revenue to fees.

Does raising my price always improve my kept-share?

In this model, yes, but with sharply diminishing returns. The jump from the $10 band to the $20 band adds about 15 points of kept-share, while later steps add roughly 4 to 5 points each. The big win is escaping the lowest band, where fixed fees dominate. Whether a price rise improves total profit also depends on how demand responds, which the model doesn't predict.

Is this based on real seller data?

No. It's a transparent model that applies Amazon's published fee structure to hypothetical products with clearly stated assumptions. It's meant to illustrate how kept-share changes by price, not to report a measured dataset. Run your own product in BeanHawk's free FBA calculator for your real numbers.

How do I find my own kept-share per SKU?

Start from your settlement reports, not your bank deposits. Sum the gross sales and every fee line Amazon charged for a SKU over a period, add your ad spend for that SKU, and divide what's left by gross sales. Doing this by hand works for a handful of SKUs; at catalog scale it's the core job of amazon fba accounting software, which should keep those per-SKU splits current automatically from each settlement.

Can I cite this report?

Yes, freely, with attribution to BeanHawk. Journalists can reach press@beanhawk.com for the underlying assumptions or commentary.