The Fee Staircase

Ask an operator what one more ounce costs to ship through Amazon and the answer is usually "a little more". That is a line: heavier costs a bit more, pricier costs a bit more, and a small change makes a small difference.

Amazon's fulfilment fee is not a line. It is flat for a stretch, then it steps up, then it is flat again. Inside a step, an extra ounce costs nothing. Across an edge, the next ounce costs the whole step. And the step is paid on every unit, every month, until the unit or the price moves.

Modeled from public data · Not a client · Not a result

What the staircase is made of

Four published things decide which step a unit stands on.

1. The size tier. Amazon sorts a unit's three sides longest to shortest and compares them with two envelopes. Small standard: up to 15 × 12 × 0.75 inches and 16 ounces. Large standard: up to 18 × 14 × 8 inches and 20 pounds. Anything bigger is oversize, which this course leaves out. A tier is not a different step: it is a different staircase.
2. The weight band. Small standard climbs in 2-ounce steps to 16 ounces. Large standard climbs in 4-ounce steps to 3 pounds, then adds $0.08 for every 4 ounces after that, before the surcharge. The weight that counts is the billable weight: the greater of what the unit weighs and its dimensional weight (lesson 3).
3. The price band. Every row has three prices: under $10, $10 to $50, and over $50. On the card in force until October 14, crossing $10 adds $0.85 to $1.05 a unit, depending on the row. Price is a step too.
4. The season. From October 15 to January 14 a second card applies. In the $10 to $50 column it runs $0.20 to $0.42 a unit higher, row for row.

On top of all four, since April 17, 2026, a 3.5% fuel and logistics surcharge is added to every fulfilment fee, the holiday card's included.

The card itself

Here are the two standard-size cards with the surcharge already in. Every fulfilment fee in this course is read from them, and so is every one in the template.

Amazon's published 2026 US FBA fulfilment fee schedules, as loaded in Hubricon's engine on September 9, 2026. Above 3 pounds, add $0.08 for each 4 ounces, before the surcharge. Amazon reprices about once a year: check the live schedule in Seller Central before you act on a figure.

Why nobody around you sees it

A P&L is an average. It tells you what fulfilment cost across everything you shipped. It cannot tell you that one SKU sits 0.8 ounces past an edge, because the cheaper version of that SKU never shipped, so it is on no report.

A cliff is only visible as a counterfactual: the same unit, one step the other way. No report anyone sells contains one. That is why the fee tool, the agency and the bookkeeper can all miss it at once. Each of them reads what happened, and the step is about what did not.

This course builds that counterfactual for any listing, from its public page, in arithmetic you can check line by line.

You can run this course on your own listings, on a competitor's, or on a listing you are thinking of launching against. Everything it needs is public.

The five numbers

Number	Where it is	What to watch
Price	The buy box.	The price customers pay most days, not a one-day deal.
Weight	Product information, "Item Weight".	A floor: Amazon bills the packed unit, which can only weigh more.
Three sides	Product information, "Package Dimensions" if shown, otherwise "Product Dimensions".	Package dimensions are closer to the box Amazon measures. Product dimensions are a floor.
Category	"Best Sellers Rank", the first line: "#2,400 in Home & Kitchen".	The top-level category, not the sub-category under it.
Sales rank	The same line: the number.	Today's rank. It moves daily; lesson 5 deals with that.

Pounds × 16 is ounces. Centimetres ÷ 2.54 is inches. Grams ÷ 28.35 is ounces.

Floors, not facts

Two of the five are floors. The page's item weight is what the product weighs; Amazon bills what the packed unit weighs, which is at least that and often more. Product dimensions stand in the same relation to the box.

A floor errs in one direction. If the page puts a unit past an edge, the packed unit is at least that far past it, so a finding built on the page understates the fee rather than inventing one.

One rule from Hubricon's engine applies here too: a size tier is never guessed. No dimensions, no tier, no finding.

Your own listings: use Amazon's measurements

For your own SKUs you don't need the public page. Seller Central's Fee Preview report (Reports, Fulfillment) lists, for every SKU, the sides and the packaged weight Amazon measured, the size tier it assigned, and the fulfilment fee it expects to charge. That is the ground truth. In lesson 8 you paste it into the Catalogue sheet, and the sheet shows you where Amazon's fee and the card disagree.

To check one listing against Amazon's own arithmetic, Amazon's Revenue Calculator prices any ASIN's fees. If it and the template disagree by more than a cent, one of the inputs is off.

Type them in

Open One listing and type the five numbers into the yellow cells. The example already there is invented; overwrite it. Leave "your own units", "landed cost" and "referral rate" blank for now: lessons 4 and 5 come back to them.

The five lines

1. Sort the sides. Longest, median, shortest. Amazon compares them in that order.
2. Billable weight. If the box is over a cubic foot (1,728 cubic inches), work out its dimensional weight: length × width × height ÷ 139, in pounds. The billable weight is the greater of that and the unit's weight. Under a cubic foot, the unit's own weight is billed.
3. The tier. Small standard if the sorted sides fit inside 15 × 12 × 0.75 inches and the billable weight is 16 ounces or less. Large standard if they fit inside 18 × 14 × 8 and it is 20 pounds or less.
4. The column. Under $10, $10 to $50, or over $50, by the price.
5. The row. The first row whose edge the billable weight does not exceed. In large standard, 8.0 ounces pays the 8-ounce row and 8.1 ounces pays the 12-ounce row. The fee is that cell, plus 3.5%.

One listing does all five, rows 18 to 31, and shows every step.

Worked through: the case study

The four edges

A unit can sit just past four kinds of edge. One listing tests all four, on both cards, and gives each a verdict.

1. The weight band. The heaviest edge below the billable weight, and how far over it the unit is. It counts within 2 ounces; further than that is a redesign, not a shave.
2. The price band. A price of $10 or more pays the $10 to $50 column; one cent under, it pays the cheaper one. Whether dropping pays depends on how far above the edge the price is (lesson 4).
3. The size tier. A large-standard unit that would be small standard but for one side: a mailer 0.9 inches thick where the limit is 0.75. It counts when exactly one side is over, by no more than 35%, and the unit weighs 16 ounces or less.
4. The box. A unit billed on dimensional weight because its box is over a cubic foot and bigger than the product needs. It counts when the dimensional weight is at least 1.3 times the real weight.

Any of the four counts only if it is worth at least 8 cents a unit. These thresholds are ours, not Amazon's: they are the ones Hubricon's engine uses to decide an edge deserves a person's attention. They sit on the Rate card sheet. Change them if your business disagrees.

Edges overlap: a lighter unit can also drop a tier. So they are never added together. The largest single edge is the finding.

An ounce lighter

gap per unit = the fee now − the fee at the edge below

Work it on both cards. For the case study that is $0.26 and $0.28.

The ounce usually comes out of the packaging: a lighter carton, a thinner insert, less void fill, a lighter bottle or jar. Sometimes it is not there at all, because Amazon measured the unit wrong. Compare the packaged weight in your Fee Preview report with what the packed unit really weighs on a scale. If Amazon's figure is higher, open a case in Seller Central and ask for the unit to be re-measured.

A smaller fill is a different product. If the ounce comes out of what the customer gets, the listing has to say so.

A cent cheaper

The cheaper price keeps more per unit, before counting any extra sales a lower price brings. Its break-even is $10.99: above that, staying over $10 pays; below it, the step costs more than the higher price earns.

It runs the other way too. A listing at $9.99 that takes "a small price rise" to $10.49 keeps $0.424 less per unit, on every sale, and no report will say why.

For any price P at or above an edge E, at referral rate r:

net per unit at one cent under = fee jump − (P − (E − $0.01)) × (1 − r)

break-even price = (E − $0.01) + fee jump ÷ (1 − r)

The same arithmetic applies at $50, where the top column starts.

A fraction of an inch thinner

Tier edges are usually the largest in this course, because they move a unit to a different staircase, not down a step.

A smaller box

What getting there costs

A gap per unit is not yet an answer. The answer is whether the gap pays back what the change costs:

months to pay back = one-time cost ÷ (gap per unit × units a month)

A new carton run, a redesign, relabelling the stock already inbound: get the quote, type it into "Is the change worth making?" on One listing, and plan on the low end of volume. The price move costs nothing to make, which is why it usually goes first.

The curve

Hubricon's engine fits a power law from rank to units. For a rank of 1,000 or worse in a top-level category:

units a month ≈ scale × 10^(5.925 − 0.875 × log10(rank))

Better than rank 1,000 it flattens, because the top of a category does not sell without limit. The scale adjusts for the category, with Home & Kitchen as the reference at 1.0; the rest are on the Rate card sheet.

Some readings from it:

• Rank 5,000 in Kitchen & Dining: about 440 a month.
• Rank 30,000 in Home & Kitchen: about 100.
• Rank 800 in Beauty & Personal Care: about 2,260.

Why it is a range

The curve is honest about its error: a factor of two either way. Ranks move every day. Categories differ in how many units a rank is worth. A listing with several variations shares one rank across all of them. A single number from this curve is a guess dressed as a fact.

So the template, like the engine, never uses it as a point. In plain arithmetic it brackets the estimate from half to one and a half times: rank 5,000 in Kitchen & Dining is 220 to 660 a month. In the simulation (lesson 6) it draws volume from a spread where one month in twenty is under half the estimate and one in twenty is over double.

A year of an edge is then the gap per unit times the units, nine months on the card in force and three on the holiday card, at the low end and the high end. One listing's "What each edge is worth" table does exactly that.

When you know the number, use it

For your own listings you don't need the curve. Business Reports, "Detail Page Sales and Traffic by Child Item", gives units ordered by ASIN. Type the monthly average into "Your own units a month" on One listing and the range closes to your number.

For a competitor, the curve is all there is. Read it as a range and say it as a range.

What is uncertain, and what is not

The fee is not uncertain: it is on the card. Three things are.

1. Volume. The rank curve's factor of two (lesson 5).
2. Cost. Freight and exchange rates move landed cost. The template lets it drift 10% either way each month.
3. The calendar. November is on the holiday card and March is not. October and January are split by the day.

How the sheet does it

Each of the 10,000 rows on the Ten thousand months sheet is one month:

• a volume draw: units × e^σz, where z is a standard normal draw and σ = ln 2 ÷ 1.645, which puts the 5th and 95th percentiles at half and double the estimate;
• a landed cost: cost × (0.9 + 0.2 × a random number between 0 and 1);
• that calendar month's fee, the two cards blended by days;
• profit = units × (price − referral − fee − cost), before ads, storage and returns;
• and what the edge is worth over a year at that row's volume.

The sheet reads off P10, the median and P90 of each, and the share of months at a loss. Recalculate (F9 in Excel, Ctrl+Shift+F9 in LibreOffice) and ten thousand new months are drawn. The percentiles barely move. That stability is the point: the range is a property of what you know, not of one lucky draw.

The case study, simulated

Modeled from public data · Not a client · Not a result

Reading a range

• P10 is the year you plan on: one year in ten comes in under it.
• P90 is the good year: one in ten beats it.
• The median is the middle of the spread, not a promise.
• The share of months at a loss is the margin's cushion: how often, at these fees and costs, a month on this listing loses money before ads.

Say the range, rounded down. "$6,400 to $19,200 a year" is a claim you can defend. A single figure to the dollar is not.

Two charges, and one of them is a cliff

Every standard-size unit in an Amazon warehouse pays monthly storage by the cubic foot: $0.78 a cubic foot from January to September, $2.40 from October to December.

Once a unit has sat 181 days, it also pays an aged-inventory surcharge, by age:

Age	Surcharge, per cubic foot a month
181 to 210 days	$0.50
211 to 240 days	$1.00
241 to 270 days	$1.50
271 to 300 days	$5.45
301 to 330 days	$5.70
331 to 365 days	$5.90
366 days and over	$6.90, or $0.15 a unit, whichever is greater

At day 271 the surcharge goes from $1.50 to $5.45 a cubic foot: 3.6 times, in one step. Amazon assesses it from a snapshot of your inventory on the 15th of each month.

Modeled from public data · Not a client · Not a result

Why the public page cannot see it

A product page doesn't say how long the units behind it have sat. Your own report does. In Seller Central, Reports, Fulfillment, Manage Inventory Health (the Inventory Age report) counts your units by age, 0 to 90, 91 to 180, 181 to 270, 271 to 365 and over 365 days, and gives Amazon's own estimate of the surcharge each band will pay. Where those estimate columns are there, use them: they are Amazon's number, not an estimate of it.

Pricing it in your numbers

1. Cubic feet per unit comes across from One listing, or type Amazon's: longest × median × shortest side from Fee Preview, ÷ 1,728.
2. Type how many units sit in each age band. The report's bands are coarser than the surcharge's, so split 181 to 270 and 271 to 365 by when each shipment arrived, or work from Amazon's estimate columns.
3. Type the units sold in the last 30 days.

The sheet prices every band, the month's total, and the cliff: how much more the units now 241 to 270 days old will cost next month if none of them sell.

The rule

days of cover = units on hand ÷ units sold a day

Amazon ages stock first in, first out. Hold more than about 270 days of cover and some units cross the cliff before they sell, at this pace, whatever else happens.

Before the 15th the choices are three: sell the oldest units through, with a price or a promotion; remove them; or liquidate them. Set the one-time cost of each against a few months of surcharge at the new rate. The sheet gives you the monthly side of that comparison; Seller Central quotes the other.

Load the catalogue

The Catalogue sheet takes 500 SKUs, one a row: price, category, weight in ounces (or pounds, in the next column), three sides, sales rank, and two optional columns: your units a month, and Amazon's own fee per unit.

For your own catalogue, the Fee Preview report gives Amazon's measured sides, its packaged weight (in pounds: use the pounds column) and the fulfilment fee it expects to charge, for every SKU. Business Reports gives units ordered. Paste them in. Every column from M onward is arithmetic, and every step of it is shown in the working columns on the right.

Check Amazon against the card

The column "Amazon's fee minus this sheet's" compares what Amazon expects to charge with what the card says for the row.

• Paste your own measurements, and it shows where Amazon measured differently. Above zero, Amazon is billing the unit as heavier or bigger than you think it is. Weigh and measure a packed unit; if Amazon's figure is wrong, open a case and ask for a re-measure.
• Paste Amazon's measurements, and it should read zero. If it does not, the card in the template is out of date or something else is changing the fee. Check it in Seller Central.

Work in order

Sort by the Rank column. It ranks every SKU by the low end of what its largest edge is worth a year. Then work down from the top:

1. Price edges first. They cost nothing to make.
2. Then packaging: weight bands, tiers and boxes, each against its quote (lesson 4).
3. Stop when a year at the low end is less than the change costs. Below that line, leave it.

When the answer is nothing

Most catalogues have a few edges worth moving and a long tail that isn't. Some have none. Of the 1,856 Amazon brands Hubricon has modeled from public pages, 1,473 showed nothing worth fixing. If your catalogue is one of them, that is a result too: the staircase is not where your money is going.

What this course cannot see

Everything above is the public half. Here is the shape of the rest, so you can size it.

• The price response. Moving a price moves volume. This course assumes volume holds, which is safe for a cut and unsafe for a rise. A real pricing decision needs that listing's demand curve, and an Amazon price test without a counterfactual is mostly noise.
• Ads. The break-even ad cost of sale on One listing is the ceiling on what an ad-driven sale can cost. Where a campaign runs above it, every sale it buys loses money. That needs your advertising reports.
• The low-inventory-level fee. Amazon adds a fee to each unit shipped when both your 30-day and 90-day historical days of supply fall under 28 days. It needs your inventory history.
• Returns and reimbursements. What comes back, what Amazon loses, and what it owes you for both. That needs your inventory ledger and returns reports.

Each is computable from your own exports, with the same discipline as this course: published cards, stated assumptions, ranges, and a counterfactual.

Run the template again every quarter, and the week Amazon publishes a new card.