True Amazon profitability on a unit is the selling price minus the referral fee, the fulfillment fee, the share of storage that unit consumed, the advertising attributed to it, the cost of returns it generated, and its landed cost. Most sellers stop after the first three, which is why their spreadsheet says 22 percent and their bank account says something else. This walks through the full calculation with a real unit, in the order the money actually leaves.
Start with the deposit, not the sale price
The single most common error is starting from list price. Start from what Amazon deposits, then work backwards. Everything Amazon charges you appears somewhere in the settlement report, and reconciling to the deposit is the only way to know you did not miss a fee category.
Pull a settlement covering a full two-week period. Do not use a calendar month. Settlement periods do not align to months, and forcing them to is how sellers double count or drop a week of fees.
The worked example
A kitchen gadget. List price $34.99, sold through FBA, in the Home and Kitchen category.
Referral fee. Per Amazon’s published selling fee schedule, Home and Kitchen carries a 15 percent referral fee, with a $0.30 minimum on most categories. On $34.99 that is $5.25. Note how much this varies by category: computers and consumer electronics sit at 8 percent, automotive at 12 percent, clothing over $20 at 17 percent, and Amazon device accessories at 45 percent. Two products at the same price in different categories are not the same product financially.
Fulfillment fee. This depends on the unit’s size tier and weight, and the rate card changes. Rather than quote a number that will be stale, pull yours from the FBA fee preview in Seller Central or Amazon’s Revenue Calculator for the specific ASIN, and date it in your model. Call it $5.40 for this example, which is in the plausible range for a small standard item but is a placeholder you must replace with your own figure.
Selling plan. The Professional plan is $39.99 a month as listed by Amazon. That is a fixed cost, not a per-unit one. Do not allocate it into unit economics. It belongs below the contribution line with the rest of your overhead, and pushing it into per-unit math is how sellers convince themselves a low-volume SKU is worse than it is.
Running total: $34.99 minus $5.25 minus $5.40 leaves $24.34 before you have paid for the product itself.
The three costs sellers skip
Storage. Monthly inventory storage is charged on cubic feet and is higher in the fourth quarter than the rest of the year. Aged inventory surcharges stack on top for units sitting past the thresholds Amazon publishes. Allocate storage per unit by taking the period’s total storage charge for a SKU and dividing by units sold in that period, not units held. A slow mover absorbs storage across very few sales, which is exactly the point you want the number to make. Assume $0.45 per unit sold here.
Returns. This is the one that quietly eats categories like apparel and electronics. A return costs you the refunded revenue, the return processing where it applies, the non-refundable portion of the original fees, and often the entire landed cost when the unit comes back unsellable. If this SKU returns at 6 percent and roughly half of returned units are not resellable, the expected cost per unit sold is meaningful even though no individual sale shows it. Model it as a percentage applied across all units rather than as an event.
Say returns cost $1.30 per unit sold once you blend refunded fees and lost inventory.
Advertising. Attribute campaign spend to the SKU it promoted, then divide by that SKU’s total units sold in the period, including organic ones. Dividing only by attributed units flatters the number badly. If you spent $2,100 on campaigns for this product and sold 1,400 units, that is $1.50 per unit.
Running total: $24.34 minus $0.45 minus $1.30 minus $1.50 leaves $21.09.
Landed cost, done properly
Landed cost is not the invoice from your supplier. It is the unit cost plus freight, plus duty and tariffs, plus customs brokerage, plus inbound shipping to the fulfillment center, plus prep and labeling, divided across the units in that shipment.
Say the factory invoice is $6.80 per unit, ocean freight and brokerage add $1.10, duty adds $0.55, and inbound plus prep adds $0.60. Landed cost is $9.05.
$21.09 minus $9.05 leaves $12.04 of contribution per unit, which is 34.4 percent of list price. That is the number worth knowing. It is also roughly a third higher than the naive calculation that stops at referral and fulfillment fees, which would have shown $15.29, and that gap is where reorder decisions go wrong.
Rules that keep the number honest
Use accrual, not cash. The unit you sold in March may have been paid for in a settlement that cleared in April, and the inventory may have been bought in January. Matching them is the entire point. IRS Publication 538 is direct about this: a taxpayer who keeps inventories generally uses an accrual method for purchases and sales, and it also sets out the permitted inventory valuation methods, including FIFO, specific identification and lower of cost or market.
Pick one COGS method and keep it. FIFO is the usual answer for imported goods with changing landed costs, because it reflects the order the inventory actually moves. Switching methods mid-year to make a quarter look better is a decision with tax consequences, not a spreadsheet preference.
Date every fee assumption. Amazon’s fee card changes, tariff rates change, and freight rates move a lot. A model with undated assumptions is a model that silently goes wrong. Write the date next to each rate and re-pull them at least quarterly.
Keep fixed costs out of unit economics. Software subscriptions, the $39.99 selling plan, your virtual assistant and your accountant all belong below contribution. Mixing them in makes every SKU look worse in a slow month and better in a fast one, for reasons that have nothing to do with the SKU.
Doing this at catalog scale
The math above takes ten minutes for one product and becomes unmanageable at 200. The blocker is not the arithmetic, it is keeping landed cost, fee allocation and return behavior current per SKU while settlements keep arriving. That is the specific problem accounting integrations for marketplace sellers are built to solve. ConnectBooks publishes a full walkthrough of profitability after every Amazon fee that is worth reading before deciding whether to automate this or keep it in a sheet.
Whichever way you go, run the calculation by hand on your three best sellers and your three worst first. You will learn more about your business in that hour than from any dashboard, and you will know whether the automated version is telling you the truth when you eventually buy one.
If you are unsure what records you are obliged to keep while you build this, the Small Business Administration’s guidance on managing business finances is a reasonable starting point.
