Amazon vs Your Own Store: Most Fee Comparisons Double-Count
- Only one fee line is genuinely a marketplace charge. Amazon’s referral fee runs 5% to 45% by category, most categories at 15%, with a $0.30 minimum. Everything else in the “Amazon takes 45%” stack — fulfilment, storage, returns handling, advertising — is a job you pay for on your own store too, just to a different payee.
- The honest comparison is referral fee against card processing plus acquisition. On a $50 item in a 15% category, Amazon’s referral fee is $7.50. The same order on your own store costs $1.75 to process at Stripe’s published 2.9% + $0.30. The difference, $5.75, is your entire per-order budget for getting that customer yourself.
- Repeat purchase moves the breakeven; fee negotiation does not. Amazon charges the referral fee on every order including repeats. You pay acquisition once per customer. At four orders per customer that same $5.75 becomes $23.00 of headroom.
- 2026 raised the fulfilment lines, not the referral line. FBA fees rose an average of $0.08 per unit on 15 January 2026, and a 3.5% fuel and logistics surcharge was added to FBA fulfilment fees from 17 April 2026 — calculated on fulfilment fees, not on the sale price.
- Amazon’s own accounts show what it is selling. In Q2 2026 third-party seller services brought in $46.78 billion, up 16%, while advertising services grew 26% to $19.81 billion. Visibility is increasingly a separate purchase from access.
Every comparison of Amazon against your own store makes the same arithmetic error. It totals Amazon’s full fee stack — referral, fulfilment, storage, returns, advertising — and sets it against an own-store figure containing payment processing and hosting. That is not a comparison, it is a scope mismatch: four of those five Amazon lines pay for work you still have to pay for when the order arrives through your own checkout, and the own-store side quietly omits them. Corrected, the decision stops being about fee percentages and becomes a single question with a number attached — can you acquire an order for less than the referral fee you would otherwise pay? This post rebuilds the stack line by line from Amazon’s published schedule, shows the arithmetic that survives the correction, works through what changed in 2026, and ends with the decision table and the operational problem that follows once you run both channels.
On this page
- What Amazon actually charges, line by line
- The double-count that breaks the comparison
- The referral fee is the only genuine marketplace charge
- Your acquisition budget is the referral fee minus card fees
- Repeat purchase moves the breakeven; negotiation does not
- What actually changed in 2026
- Amazon’s own numbers show what it is really selling
- A decision table you can apply this week
- Running both is the usual answer, and it is an inventory problem
- When this analysis does not apply
- What to pull before you decide
What Amazon actually charges, line by line
Amazon’s seller costs are five separate charges with five separate rate cards, and conflating them is where most analysis goes wrong. The referral fee is a percentage of the sale price, published at 5% to 45% depending on category with most categories at 15% and a $0.30 minimum. Fulfilment by Amazon fees are per-unit and driven by size tier and weight. Storage is charged per cubic foot per month and rises in the fourth quarter. Returns processing is its own rate card. Sponsored Products is a bid auction you opt into. On top of the selling fees sits the plan itself: $39.99 per month for a Professional account, or $0.99 per item sold on an Individual account.
Only the first of those is set purely by Amazon and unavoidable on a completed sale. The rest vary with your product’s dimensions, your inventory turn, your return rate and your bidding. Any article quoting you a single “Amazon takes X%” figure has averaged across categories and product shapes that have nothing to do with yours.
| Charge | Calculated on | Does your own store avoid it? |
|---|---|---|
| Referral fee (5–45%, most 15%, $0.30 minimum) | Sale price | Yes — no equivalent charge exists |
| FBA fulfilment fee, plus 3.5% surcharge | Unit size tier and weight | No — a 3PL invoice or your own pick-and-pack labour |
| Monthly inventory storage | Cubic feet, seasonally rated | No — warehouse rent or 3PL storage |
| Returns processing | Per returned unit, by category | No — your own returns handling cost |
| Sponsored Products | Your bids, optional | No — paid search, paid social, SEO or affiliate |
| Selling plan ($39.99/month or $0.99/item) | Flat | Partly — replaced by hosting and platform costs |
Verdict: one row in that table has no own-store counterpart. The other five describe work that has to happen and be paid for regardless of which channel took the order. That single observation is the whole argument, and the rest of this post is its consequences.
The double-count that breaks the comparison
The standard comparison double-counts because it prices Amazon at full scope and your own store at partial scope. Set out as a mapping, the asymmetry is obvious: five cost functions exist in both channels, and only one of them is structurally different.
Read the bracket first. Fulfilment, storage, returns and advertising are not marketplace charges — they are the cost of running a physical product business, and they follow the order wherever it arrives. Moving a sale off Amazon does not delete them, it re-addresses the invoice. Only the top row disappears, and only the top row should appear in a channel comparison.
A fee comparison invites you to negotiate or optimise fees, and none of the five lines are meaningfully negotiable for a mid-market seller. A scope-corrected comparison points at demand acquisition, which you can genuinely change. The teams that leave Amazon successfully do it because they built a cheaper way to be found, not because they found a cheaper fee.
The referral fee is the only genuine marketplace charge
Strip the shared rows and one line remains: the referral fee, published at 5% to 45% of the sale price by category, with most categories at 15% and a minimum of $0.30. That charge buys three things bundled together — a listing in a catalogue buyers already search, a checkout they already trust, and card processing, for which Amazon’s seller fee schedule carries no separate line. It is contingent, in that no sale means no fee, and it is charged on the gross sale price rather than on your margin.
Your own store buys the same bundle differently. Payment processing is a published, predictable rate: Stripe lists 2.9% + $0.30 for domestic online card charges, with an additional 1.5% for international cards and a further 1% where currency conversion is required. Trust and catalogue presence are not purchasable at a rate card; they are built with time and spend. So the referral fee is best read not as a tax but as a price quote for demand — one you can accept, or undercut by building your own.
That reframing is what makes the comparison tractable. Two published rates, one subtraction, and the result is a budget.
Your acquisition budget is the referral fee minus card fees
Take a $50 item in a 15% referral category. Amazon’s referral fee on that sale is $7.50. The same $50 order taken through your own checkout costs $1.45 in percentage fees plus a $0.30 fixed fee at Stripe’s published rate, or $1.75 in total. The gap is $5.75, and that is the exact amount you can spend acquiring that order before your own store is worse off than the marketplace.
Nothing else belongs in the calculation, because everything else is common to both channels. If you pick and pack it yourself either way, the pick-and-pack cost cancels. If it sits in the same warehouse either way, the storage cost cancels. The subtraction is deliberately small, and its smallness is the point: $5.75 per order is a far tighter constraint than the “Amazon takes 40% versus your 5%” framing implies. Sellers who make the naive comparison budget as though they have $17 of headroom, buy traffic accordingly, and discover the shortfall two quarters later.
Substitute your own numbers rather than borrowing this example. Your referral rate depends on your category, and your average order value moves the fixed $0.30 component around — at a $15 average order value the fixed fee is 2% of the sale on its own, and at $200 it is negligible.
Repeat purchase moves the breakeven; negotiation does not
The single-order comparison understates your own store, and the reason is structural rather than promotional. Amazon charges the referral fee on every order, including the fifth order from a customer who has bought from you four times already. Acquisition on your own store is paid once per customer and then amortises across everything that customer buys afterwards. Repeat purchase is therefore the variable that decides the channel question, and it is the one variable most fee comparisons never mention.
Find your own repeat rate on the left axis and read the budget across. A brand selling a consumable at three orders per customer per year has roughly triple the acquisition headroom of a brand selling a once-in-a-decade purchase, on identical fee rates. This is why the same channel advice produces opposite outcomes for a coffee subscription and a mattress, and why category-level generalisations about marketplace fees are close to useless.
It also explains a pattern worth naming: the products best suited to leaving Amazon are rarely the ones with the worst fee rates. They are the ones with the highest repeat rate, because repeat purchase is the only mechanism that compounds a fixed acquisition cost into a widening advantage.
What actually changed in 2026
Two changes landed this year, and both hit the shared rows rather than the referral row — which means neither changes the breakeven arithmetic above, and both change the fulfilment comparison. Amazon’s 2026 fee update took effect on 15 January 2026 and raised FBA fees by an average of $0.08 per unit sold, which Amazon characterises as less than 0.5% of an average item’s selling price. Amazon also stated there would be no new FBA fee types in 2026, following no increase to US referral and FBA fees in 2025.
The larger change came in April. Amazon added a 3.5% fuel and logistics-related surcharge to fulfilment fees, effective 17 April 2026 for FBA in the US and Canada and for Remote Fulfilment with FBA into Canada, Mexico and Brazil, and from 2 May 2026 for Buy with Prime and Multi-Channel Fulfilment. The surcharge is calculated on fulfilment fees rather than on the sale price, and Amazon put the average impact at about 17 cents per unit. No end date was published.
| 2026 change | Effective | Applies to | Effect on the channel decision |
|---|---|---|---|
| FBA fee update, +$0.08 per unit average | 15 January 2026 | FBA fulfilment fees | Fulfilment comparison only — referral rate unchanged |
| 3.5% fuel and logistics surcharge, ~$0.17 per unit | 17 April 2026 (FBA, US and Canada) | Fulfilment fees, not sale price | Fulfilment comparison only — no end date given |
| Same surcharge extended | 2 May 2026 | Buy with Prime, Multi-Channel Fulfilment | Raises the cost of using Amazon logistics for your own store’s orders |
| Referral fee rates | Unchanged | Sale price | The breakeven calculation above is unaffected |
Verdict: the third row is the one to watch if you fulfil your own store’s orders out of Amazon’s network. Multi-Channel Fulfilment and Buy with Prime are the bridge many sellers use to run one inventory pool, and the surcharge now applies to that bridge — which shifts the make-or-buy decision on logistics without touching the marketplace question at all.
Amazon’s own numbers show what it is really selling
Amazon’s quarterly disclosures answer a question its fee schedule does not: whether the referral fee still buys demand, or merely buys access. According to Amazon’s second-quarter 2026 results, third-party seller services net sales reached $46.78 billion, up 16% from $40.35 billion a year earlier, while advertising services grew 26% to $19.81 billion from $15.69 billion. Together those two lines came to $66.59 billion against total net sales of $200.61 billion — almost exactly a third of the company.
The growth differential is the part sellers should read closely. Advertising is growing roughly ten percentage points faster than seller services, which means the share of marketplace revenue that comes from sellers paying for visibility is rising against the share that comes from sellers paying for transactions. Put plainly: the referral fee increasingly buys a listing, and being found on that listing is a second purchase.
That trend erodes one side of the trade this whole post rests on. The referral fee is defensible when it delivers demand you could not otherwise reach. If reaching that demand requires a competitive Sponsored Products bid on top, then part of Amazon’s demand advantage has already been converted into a variable cost that looks a lot like your own advertising spend — and the comparison tilts further toward whoever has the cheaper route to being found.
A decision table you can apply this week
The corrected arithmetic produces conditions rather than a verdict. Find the row that matches your situation, and treat the recommendation as the default you would need a specific reason to override.
| If this is true of your product | Default channel | Reasoning |
|---|---|---|
| Blended acquisition cost per order exceeds the referral fee minus card fees | Marketplace | You are paying more to be found than the fee you are trying to avoid |
| Repeat orders per customer above roughly two, with a known reorder trigger | Own store | Acquisition amortises; the referral fee does not |
| Low differentiation, buyers search by product type rather than brand | Marketplace | Own-store demand has to be bought outright, order by order |
| Strong brand search volume already exists for your name | Own store | The demand is already yours; the referral fee buys nothing you lack |
| Category referral rate at the high end of the 5–45% range | Own store | Headroom per order is wide enough to fund real acquisition |
| Average order value under about $15 | Marketplace | The $0.30 fixed card fee and per-order acquisition both bite hardest here |
Verdict: most mid-market sellers match rows in both halves of that table, which is why the honest answer for most catalogues is not one channel but a split — high-repeat, high-differentiation lines direct, and discovery-driven or low-repeat lines on the marketplace.
Running both is the usual answer, and it is an inventory problem
Running both channels is a defensible outcome of the analysis above, and it converts a commercial decision into an operational one. One inventory pool now serves two demand sources that cannot see each other, and the failure mode is specific: both channels accept an order against the same unit, and neither was wrong at the moment it accepted. The cost of that failure is asymmetric — a cancelled marketplace order damages seller metrics in a way a cancelled own-store order does not.
The mechanics are covered in detail in our guide to keeping WooCommerce, FBA and NetSuite in sync without overselling, and the underlying choice of which system owns the stock number is set out in inventory sync direction, one-way versus two-way. The timing question — whether a channel learns about a stock change by event or by poll — decides how wide your oversell window is, and is worked through in webhooks versus polling for inventory sync. If orders can ship from more than one location, settle multi-warehouse order routing before adding the second channel rather than after.
Three rules keep a two-channel operation honest. Designate one system as the authoritative owner of available-to-sell, in writing, and let every channel read from it rather than hold its own count. Hold a buffer against the marketplace listing sized to your sync interval, because the buffer is cheaper than the metric damage from a cancellation. And reconcile on a schedule against physical count, because two-channel drift accumulates silently and is invisible in both channels’ own reports. The wider integration sequence sits in the NetSuite and WooCommerce integration guide library, and the same channel-evaluation logic applied to a newer marketplace is in our analysis of whether TikTok Shop is worth a merchant’s time.
When this analysis does not apply
The subtraction at the centre of this post assumes the two channels are selling the same unit to the same buyer, and there are cases where that assumption fails badly enough to invalidate the conclusion.
The clearest is contingency. Amazon’s referral fee is charged only on a completed sale, whereas advertising spend on your own store buys a chance at a sale and is spent whether or not one happens. At low volume, or where conversion is unproven, that asymmetry is worth real money and favours the marketplace beyond what the arithmetic shows. A seller with $5.75 of headroom per order and no reliable channel to spend it in does not have $5.75 of value.
Three further cases: products where the marketplace is genuinely the search engine, so own-store demand would have to be created rather than captured; regulated or high-consideration categories where the marketplace listing format cannot carry the information needed to sell; and businesses whose working capital cannot absorb the timing difference between paying acquisition up front and recovering it across future orders. Each of those makes repeat-rate amortisation theoretical rather than bankable.
Finally, none of this is a case for leaving a channel that works. Removing a profitable marketplace listing to protect a margin percentage lowers total contribution, and contribution pays salaries where percentages do not.
What to pull before you decide
The analysis needs six inputs, all of which you already have. Work down the list before modelling anything, because five of the six are commonly estimated when they could be measured.
- Confirm your category’s referral rate against Amazon’s current published schedule rather than a rate quoted in an article — the published range spans 5% to 45%.
- Pull your effective card rate from your processor’s statement, not the headline rate, so international card and currency-conversion loading is included.
- Calculate orders per customer over a rolling twelve months, by product line rather than in aggregate, because catalogue averages hide the lines that justify a direct channel.
- Measure blended acquisition cost per order across every paid channel, including the fixed costs of running them, not just media spend.
- Separate fulfilment and storage costs from marketplace fees in your own reporting, so the shared rows stop contaminating the comparison.
- Check whether Multi-Channel Fulfilment or Buy with Prime sits in your own store’s fulfilment path, since the 3.5% surcharge reached both on 2 May 2026.
Run those six against the decision table above and the answer usually resolves per product line rather than for the business as a whole. That result is inconvenient operationally and correct commercially, and it is the point at which the WooCommerce and NetSuite integration checklist becomes the practical next document.
Get the working checklists
The runbooks and decision checklists from these guides, as printable PDFs — free in the SoftXone guide library.
If the decision lands on running both channels, the work that follows is a single authoritative stock number serving two systems that will never agree on their own. That is scoped as a multi-channel inventory integration rather than a listing exercise: one owner for available-to-sell, an agreed sync interval, a buffer sized to it, and reconciliation that runs whether or not anyone is watching.
References
- Amazon — Selling plans and feesAmazon’s published pricing: referral fee range of 5% to 45% by category, $0.30 minimum referral fee, $39.99 Professional plan and $0.99 per item Individual plan.
- Amazon — Update to U.S. referral and Fulfillment by Amazon fees for 2026Amazon Selling Partners: average $0.08 per-unit FBA increase effective 15 January 2026, no new FBA fee types in 2026, no 2025 increase.
- Amazon to apply 3.5% fuel and logistics surcharge on fulfillmentSupply Chain Dive — reporting Amazon’s April 2026 seller notice: 3.5% calculated on fulfilment fees not sale price, effective 17 April for FBA and 2 May for Buy with Prime and MCF, about 17 cents per unit, no end date.
- Amazon.com Announces Second Quarter ResultsAmazon investor relations — Q2 2026 net sales disaggregation: third-party seller services $46,780M (+16%), advertising services $19,809M (+26%), total net sales $200,606M.
- Stripe — PricingPublished card rates used in the breakeven calculation: 2.9% + $0.30 for domestic online card charges, plus 1.5% for international cards and 1% where currency conversion applies.
Frequently asked questions
How do I find the exact FBA fulfilment fee for my product?
Amazon’s Revenue Calculator returns a per-ASIN figure derived from your product’s size tier and weight, and Amazon labels those results as estimates. Pull it per ASIN rather than per category, because size tier and weight move the fulfilment fee far more than category does. Any figure captured before 2026 is stale on two counts: the fee update that took effect on 15 January 2026, and the 3.5% fuel and logistics surcharge added on 17 April 2026.
Does the 3.5% surcharge apply to the referral fee as well?
No. Amazon calculates the surcharge on fulfilment fees, not on the sale price, so it scales with your product’s size and weight rather than with what you charge for it. A higher-priced item in the same size tier therefore absorbs it more easily. The surcharge also reached Buy with Prime and Multi-Channel Fulfilment on 2 May 2026, which means own-store orders shipped from Amazon’s network carry it too. Amazon published no end date.
What repeat purchase rate justifies building a direct channel?
There is no universal threshold, because it depends on your category’s referral rate and your average order value. Compute headroom per order as the referral fee minus your effective card processing rate, multiply by orders per customer over a rolling twelve months, and compare the result against blended acquisition cost per order. Measure it per product line rather than across the catalogue, because catalogue averages hide the specific lines that clear the bar.
Can I cut the effective rate by switching off Sponsored Products?
Sponsored Products is an opt-in auction, so it can be switched off. Whether the sales survive is the real question, and the trend is not encouraging: Amazon’s advertising services revenue grew 26% year on year in the second quarter of 2026 against 16% for third-party seller services, which indicates competition for visibility is intensifying rather than easing. Test it by product rather than switching it off across the account.
Is Multi-Channel Fulfilment cheaper than a separate third-party logistics provider?
It removes a second warehouse, a second integration and a second process, which is why sellers running both channels reach for it. Since 2 May 2026 it also carries the 3.5% fuel and logistics surcharge, so any comparison against a standalone provider made before that date needs redoing. Price both against your actual order profile — weight, size tier and destination mix — rather than against headline rates.
Leave a Reply