Why Your Estimated FBA Fee Differs From Amazon's Quote (2026)
A fee estimate is either a lie with two decimal places or a number you understand the error bars on. The second kind is more useful and harder to build, because it requires knowing exactly where it will be wrong.
Here is where this one is wrong: six mechanisms, each of which turns a $4.15 line into a different number on your invoice — and the design choices that keep the difference visible instead of hidden.
Where the numbers come from
No Amazon API is involved, and that shapes everything. The extension asks for two browser permissions — local storage and a timer — and no seller credentials. So the fee schedule cannot come from your account; it comes from the public rate pages, transcribed by hand into a table that lives in the codebase.
That transcription is versioned and dated: the current table carries version 2026.11, dated 2026-11-01, and names the Seller Central reference page it was copied from. The version is not buried — it is rendered in the panel next to the numbers, so you can tell which schedule produced the figure in front of you. A stale fee table is a silent failure, and the only honest defence is to make its age visible.
How the table stays current
Amazon changes fees several times a year, and an extension that ships its rates in the package cannot wait for a store review to keep up. So the table is split from the build:
- A bundled copy ships inside the extension and is always available, offline, unchanged.
- A background timer fires once a day and fetches a JSON file from the extension's own website.
- The response is accepted only if its schema version matches what this build expects and its fee-table version is newer than the installed one. Anything else is dropped.
- The accepted copy is stored locally and takes precedence over the bundled table on the next calculation. A failed fetch is not an error message — the bundled table simply keeps working.
The shape of that design tells you its limits: the update path can fix a rate change within a day of the schedule being transcribed, but it cannot tell you anything about your product's measured weight. That gap is the rest of this post.
Reason 1: Amazon measures, you type
The fulfilment fee is charged on what the warehouse finds, not what the listing says. Two units of the same SKU arrive with different cartons, and the heavier one is billed as the heavier one.
- Weight. The bands in the table are upper bounds: 4 oz, 8 oz, 12 oz, 16 oz. A unit that measures 16.2 oz does not pay the 16 oz price.
- Dimensions. All three sides matter, and size tiers are tested on the sorted dimensions — the longest side against the tier's longest limit, and so on. Packaging counts, not just the product.
- Dimensional weight. Above the standard sizes, billed weight is the greater of actual and volume-based weight. A light, bulky item is charged as if it were heavy.
Where the page publishes no dimensions, the calculator assumes a common shape — roughly a one-pound standard-size unit — and says so in the toolbar. That assumption is the largest single source of surprise on a search-results badge, which is precisely why the badge is not the place to make a sourcing decision.
Reason 2: tier cliffs, not slopes
Inside a tier, price moves in steps. Between tiers, it jumps. The US schedule has four tiers:
| Tier | Max dimensions | Max weight | Fee shape |
|---|---|---|---|
| Small standard | 15 × 12 × 0.75 in | 16 oz | $3.06 → $3.60 in bands |
| Large standard | 18 × 14 × 8 in | 21 lb | $3.68 → $6.17 in bands, then base + per half pound |
| Large bulky | 59 × 33 × 33 in | 50 lb | $9.61 flat + $0.38 per lb over |
| Extra-large | 59 × 33 × 33 in | 150 lb | $26.33 flat + $0.38 per lb over |
The cliff is the thing to remember. The 0.75-inch thickness limit means a product and its slightly thicker cousin — the same retail price, the same weight, one extra millimetre of blister pack — leave the small-standard table and enter the large-standard one, where a 12 oz unit costs $4.15 instead of $3.24. Packaging design is a margin decision, and the tiers are where it bites.
Apparel pays its own small cliff: a per-unit surcharge on top of the tier fee, because garments are handled differently. Getting that flag right is a manual switch, because the page does not always say.
Reason 3: category and price exceptions
The referral rate is a category lookup, and the category is whatever Amazon has classified the ASIN as — which is sometimes not what you would call the product. Rates in the table run from 8% to 45%, with 15% as the default and the interesting ones (apparel 17%, jewellery 20%, watches 16%, Amazon device accessories 45%) living far from the default.
Below a $10 price, a few categories — beauty, health, grocery, pet, baby — drop to 8%. The rule reads like a subsidy for cheap consumables and is missed by every calculator that treats referral as one number.
Both mechanisms produce the same class of error: the estimate uses a plausible rate while the invoice uses the real one. On a 15% versus 45% mistake, no amount of arithmetic downstream saves you.
Reason 4: storage is seasonal and partially unmodelled
Storage is cubic feet × rate × months, and the standard-size rate in the US is $0.87 in the low months and $2.40 from October through December. Ask a March question about a November inventory position and you are off by a factor of about three on that line — which is why the month is an input rather than an assumption.
Being straight about one simplification: the schedule also lists a lower oversize storage rate, but the calculation applies the standard row to every unit. On an oversized SKU kept for months, that single choice can move the storage line enough to matter — and it is the kind of limitation that only exists for anyone who reads the code, which is the reason this paragraph is here.
Reasons 5 and 6: averages and omissions
Inbound placement is an average. The US figure used is $0.35 per unit, which is a weighted guess across minimum-placement and consolidated-choice options, regions and box counts. A real quote requires a shipment plan; the toggle exists so you can leave it out of a comparison where it does not apply.
Some fees are not modelled at all. The refund administration fee — the lesser of $5 or 20% of the referral fee on a refunded order — sits in the table but is not applied by the profit line; returns are modelled as a percentage of price with nothing recovered. Long-term storage surcharges, aged inventory, low-inventory and removal fees are entirely outside a per-unit margin. Each omission is conservative in the same direction: they can only make the real picture worse than the shown one, never better.
Using a number you cannot fully trust
The practical answer is to treat an estimate as a band and promote it to a number only when you have evidence:
- Use the panel to filter hundreds of candidates down to a handful — that is what the search badges are for.
- On the survivors, override the fulfilment fee with a quote from Amazon's own revenue calculator, or with the invoice from a unit you have already shipped.
- Keep the storage months honest: a slow-moving SKU held 180 days is not a one-month calculation.
- Compare the tier you assumed against the one you got. If the estimate kept being wrong in the same direction, that is a modelling habit, not bad luck.
The useful property of a per-line breakdown is that you can see which line the disagreement lives on. A single profit number tells you nothing about whether the model or the world changed.
FAQ
Does an FBA fee calculator use Amazon's API?
This one does not, and most browser extensions like it cannot. There is no seller identity, no signed request and no report download; the numbers come from a fee schedule transcribed into a table that ships inside the extension. That is the whole reason the word estimate is attached to every line rather than buried in a disclaimer.
How often is the fee table refreshed?
Once a day. A background timer fetches a versioned JSON file from the extension's own website, checks that its schema matches the build it is talking to, and keeps it only if its version is newer than the table already installed. The bundled copy stays as the fallback, so a failed update changes nothing you can see.
Why is my actual fulfilment fee higher than the estimate on the same product?
Usually because the fulfilment centre measured something differently. Weight and dimensions are re-measured on receipt, and a unit that exceeds a size limit by a fraction crosses into the next tier's whole price band rather than paying a little more. Category classification and any dimensional-weight rule on the size tier move it in the same direction.
Should I override the fee if I disagree with it?
Yes, whenever you have a better number — a quote from the revenue calculator, an invoice from a shipped unit, or a size tier you know applies. The override is designed for exactly that, and it replaces only the line you touch, leaving the rest of the chain computed. The result is then as good as your input, which is the honest trade.
More guides in the VKT blog; questions to [email protected].
