How Seller Profit Doctor works out your numbers, and what it cannot know

The check-up takes what you tell it, such as orders a month, prices, what the product costs you, shipping, ads and your plan, and applies each platform’s published fee rules to it. The profit, the fees and the losses it finds are arithmetic: written formulas that give the same result for the same answers, the same fee rates and the same day. This page describes the free check-up and the paid report. Our Shopify app, which reads a store’s own orders, has its own page.

Arithmetic and advice are separate. What your shop keeps, what a fee costs and what a loss adds up to are calculations. A fix is a what-if: it says what a change would be worth at your numbers, if orders hold, and shows how far orders could fall before the change would leave you worse off. It is not a prediction.

No AI works out a number. The check-up and the paid report run no AI. Ask Doc, the AI assistant on /ask, explains fees and can run the same calculators. A check in our code requires each figure in its answer (small counts and years aside) to appear in what the calculators returned or in your own message; the answer is corrected once, and a sentence that still carries any other figure is removed. How our AI tools work is on its own page.

It is not an accounting audit. The check-up does not log in to your shop, read your statements or see your bank. It estimates from the answers you give, so its figures are only as good as those answers.

What your shop keeps in a month is money in minus money out. Money in is what buyers pay for the item and for shipping, before any platform takes its share. Money out is everything the check-up counts: what the product costs you, packaging, the shipping labels you buy, the platform’s fees, ads, and the plan you pay for.

• Etsy. Item sales and shipping buyers pay, less materials, packaging, labels, the listing, transaction and payment-processing fees, Offsite Ads and Etsy Ads.
• Shopify. Orders and shipping buyers pay, less card fees, product cost, labels, ads and your Shopify plan. Packaging is counted only if you give it.
• Amazon. Unit sales, less referral fees, fulfillment (FBA) fees, your selling plan, landed cost, and Sponsored Products.

Per order and per month. The engine works out one order’s fees line by line, rounds each line to the cent, and multiplies by the number of orders, which is why the receipts below read “100 × $3.78”. Every amount is whole cents, there is one rounding rule (half a cent rounds up, after floating-point noise is removed), and every total is the sum of its rounded lines, so a receipt always adds up. Etsy does not publish how it rounds each fee line; we assume it rounds to the cent. Amounts shown in whole dollars are for reading only and are never fed back into the math.

Revenue is not profit. The Etsy example below takes in $3,500.00 in a month and keeps $1,180.00. Sales, money in and what you keep are three different numbers, and the check-up reports the last.

What is left out. Anything the check-up does not ask about is not in the number: rent, other software, income tax and similar costs. On Etsy and Shopify, sales tax is left out of the fee math unless you give it. Your own time is not a cost in the receipt: if you give it, the report shows what an hour of work earned you, and it never counts toward a loss or the grade.

Each example is one fictional shop, worked out by the same engine, with the same functions that produce your result, on an explicit day: October 13, 2026. The day is fixed in the page’s code, so the examples do not change when a fee season starts. Amazon’s holiday peak window opens on October 15, 2026, two days later; the 30-day comparison below shows what that does to the same seller. You can see the same shops, with their fixes, on the example result pages.

What is assumed in an example is tagged Assumed, and what rests on a third-party rate is tagged Estimate. A real seller replaces the assumed lines with their own answers.

The grade is one of five words, set by two written rules. The worse of the two wins. The lines are our own rule, set by us; they are not an industry benchmark, and the grade is not a profit margin or a ranking against other shops.

Only losses we are sure of count, including sure losses that rest on third-party estimates (the grade says how much of it is estimated). “Up to” amounts, checks that could not run, and your own pay never count. Comparisons are made in whole cents, and a figure exactly on a line falls in the band that starts there: losing exactly 3% of sales is the second band, and keeping exactly 20% of an order is not capped.

Not enough information. A grade needs at least 3 checks to have run, and at least half of the checks counted for your shop (for a shop with 8 checks, 4). Until then the result says “Not enough answers yet” and names the answers that would let more checks run. Losses found still show.

Every number in a result is one of these, and the check-up says which.

• Entered by you. The answers you give. If a band was chosen (“76 to 150 orders”), the engine uses a stated figure and lists it.
• Official. A fee published on the platform’s own pages, or on a tax authority’s or carrier’s. Each has a source link, the source’s own date and the day we checked it.
• Assumed default. A number we use until you give yours, such as packaging per order or the share of sales from Offsite Ads. It is not a published fee, whatever its source. Results list these under “What we assumed” with a link to change each one.
• Third-party estimate. A figure from a source that is not the platform, used where the platform keeps its table behind a login (most of Amazon’s fulfillment fees). Every figure built on one is marked Estimate, kept beside sure dollars, and counted in the grade.
• Not confirmed. A figure no primary source backs. It never counts as a sure loss: it appears as a note, or as an “up to” amount that stays out of the loss total and the grade.

Provenance (where a figure was published), verification (whether we could confirm it) and assumption (whether we picked it) are different facts, and a figure can be more than one: the card rate we assume for another processor is Stripe’s real published rate, and it is still an assumption about your processor. The badge on /rates puts an assumption first.

Uncertainty, for example. The 10% share of sales we assume comes from Offsite Ads is the figure a 2020 news article quoted from Etsy’s launch. It is an average across the platform, not your shop, and sellers report from about 2% to 20%. It is labeled a third-party figure and an assumed default, and your result lets you change it.

Some fees depend on the date. Amazon charges a holiday peak amount per unit from October 15, 2026 to January 14, 2027, both days included, and its storage rate is higher from October through December. Etsy and Shopify results mention the USPS surcharge only inside its window (April 26, 2026 to January 17, 2027); it changes a note, not the profit figures.

One calendar. The day is the calendar day in Los Angeles. The free check-up, the AI assistants’ check-up and the paid report each read the clock once and give the engine that same day, so on a given day they agree. A result that names its own day, as the examples above do, keeps it.

A result can move when you did nothing. For an Amazon seller, the same answers worked out the day before October 15, 2026 and on it give different results, because the peak fee starts that day. That is the platform’s schedule, not a change in your shop.

Reports are not frozen snapshots. The paid report page works out its figures again each time it is opened, on that day, with the fee rates the site holds that day. The 30-day comparison is the part that deliberately works the first answers out on the day the report was paid for, and again on the re-check day, so the two can be told apart.

The full report includes a re-check 30 days after it was paid for: you answer the same questions again. A fee that changed in between would otherwise look like a change in how your shop is doing, so the comparison works the numbers out three ways and splits the difference:

• Your own change. Your new answers against your first answers, both under the fees of the re-check day.
• The fee schedule. Your first answers on the re-check day against the same answers on the first day. Only the platform’s dated rules differ.
• The actual change. Your new answers on the re-check day against your first answers on the first day. It is your own change plus the fee schedule’s, to the cent.

The comparison shows what moved, not why. A change in your own numbers can come from a fix you made, from the orders you got or from your answers, and it does not say which of those is the reason. When both checks fall on the same side of every fee date, the fee schedule’s effect is $0.00.

The numbers are checked by automated tests that run in our code repository. These are the ones that pin down the calculations, by what each one checks. The rest of the test suite (pages, payments, emails, storage) does not check fee accuracy, and is not listed.

• Whole cents and one rounding rule

  Typed dollars become exact cents, half a cent rounds away from zero, and floating-point noise cannot flip a result.

• Worked shops reproduced by hand

  The Etsy and Shopify example shops are worked out by hand in their source files, and the tests check the engine returns those figures. On all three platforms the tests check that a month’s lines add up exactly to what the shop keeps.

• Fee tables at their edges

  Amazon’s referral bands are tested on both sides of each price line ($20.00 itself is in the lower band), including categories that charge each rate on its own slice of the price.

• Grade thresholds

  The loss-share lines (3, 7, 15 and 25 percent) and the kept-per-order lines (20, 10 and 0 percent) are tested one cent either side, and so is the rule that the worse of the two wins and the rule for when a grade is withheld.

• Fee dates and the Los Angeles day

  Amazon’s peak window is tested on its first and last day and the days around them, at times when it is already tomorrow in UTC but still today in Los Angeles, and with the USPS surcharge window on Etsy and Shopify.

• Free, paid and AI check-ups agree

  On the same day, the free result, the paid report and the AI assistants’ check-up give the same kept profit per order and per month on every platform, and Etsy and Shopify money does not move with the date.

• The 30-day comparison

  Your own change plus the fee schedule’s effect equals the actual change, to the cent, for every figure and on every platform. Identical answers never show as a change in your numbers.

• Source labels on the rates page

  A row is labeled “Official” only when its source is the platform, a tax authority, a carrier or the payment processor itself, and a default we assume is never labeled “Official”.

This page is tested too: every figure in its examples is compared with the engine and with arithmetic written out in the test.

What this is not. These are our own tests of our own code. They are not a third-party audit or certification. They check that the code does what its written rules say; they cannot tell us that a platform changed a fee today. That is what the dated source checks on /rates are for.

• Your answers. A result is only as accurate as what you enter. Ranges and “not sure” become a stated figure, and some costs use defaults or estimates until you replace them.
• Your account. Marketplace charges vary with account, category, location and plan. These tables cover US sellers in US dollars. On Etsy and Shopify, sales tax is left out of the fee math unless you give it, and every Shopify order is treated as paid by card online unless you say otherwise.
• Amazon. Fulfillment, storage, returns and other fees depend on your product’s size and weight, your inventory and your account, and Amazon keeps its tables behind the seller login. We use third-party rate cards and mark every dollar built on them. Amazon’s Fee Preview shows your real fees. It is also not confirmed whether Amazon’s price for the referral fee includes shipping on orders you ship yourself; we assume it does.
• Fee changes. Fee schedules change, sometimes without notice. We check each platform again Etsy every six months; Shopify every six months; Amazon every three months, and sooner if a platform announces a change.
• What a fix is worth. The amount shown is arithmetic on your answers, if orders hold. It is not a promise of results, and yours can differ.
• Your books. A result does not replace your actual statements or professional accounting or tax advice.
• Recalculation. A report page is worked out again each time it opens, with the engine and the fee rates as they are that day, so a figure in a report you opened last month can differ from the one you see now.

Every rate is on the fee rates page with its source link, the source’s own date, the day we checked it and how exact it is. Where a platform’s page could not be read directly, the table says so, and a reading through the Internet Archive is linked as such.

• Etsy: checked October 13, 2026; next check by April 13, 2027.
• Shopify: checked October 13, 2026; next check by April 13, 2027.
• Amazon: checked October 13, 2026; next check by January 13, 2027.

Last reviewed. This page was last read against the code and the sources on October 18, 2026.

Report a correction. If a figure, a source or a sentence here looks wrong to you, write to support@sellerprofitdoctor.com with the page, the figure and where you see it differently. A link to the platform’s page helps. A confirmed correction goes into the history below.