Your marketplace dashboard shows revenue. Mine did too — right up until I sat down and subtracted, by hand, everything the dashboard leaves out.
The listing that taught me the lesson is the one I use as this tool's sample. A ceramic mug: 60 sold, the best volume in the shop. Here's what the honest subtraction looks like:
- Kept gross after 9 returns: 841.50
- Fees: listing 0.20 · transaction 54.70 · processing 37.99 · offsite ads 100.98 = 193.87
- Materials (COGS): 346.80
- Shipping I absorbed, on all 60 units shipped including the returns: 534.00
- True net: −233.17. Margin −27.7%. Per unit sold: −3.89.
The bestseller was the biggest loser in the catalogue. Every SEO tool in the world would have helped me rank it higher.
So I built the subtraction into a small CLI: seller-profit-truth. It reads a JSON or CSV file of your listings (Etsy and Gumroad), and prints per listing: gross kept, every fee line separately, COGS, shipping cost, true net, true margin — and a LOSS-MAKER flag, with exit code 1, so a script can catch a bleeding catalogue without parsing prose.
Watch it run (27 seconds): the demo video runs the tool on the bundled sample — the mug flagged LOSS-MAKER at −233.17, the catalogue totals, then a break-even run on the mug's own numbers (minimum price: 18.41). Watch the demo
It also solves the inverse problem. breakeven answers: what must this cost to truly keep my target margin after fees? My hand-poured candle sells at 14.00 and keeps 1.5% (30 sales → 6.06 of profit). For a true 20% margin it has to cost 17.04 — a number no dashboard will ever show you.
The model, and why it's stated out loud
Every profit tool makes quiet choices. This one prints its choices in every report:
- Kept units = sold − returns. A refunded unit hands the money back; it's not revenue.
- Fees are charged on kept units only (platforms credit their fees on refunded sales).
- Shipping cost counts on every unit shipped, returns included. You paid to ship the returned mug. That money is gone.
- COGS on kept units only — a returned unit is restocked.
- Etsy's listing fee is per listing and owed even at zero sales.
The honesty choices (the part I care about most)
Fee rates are defaults, stamped, never "facts". The encoded rates are stamped "as of 2026-11" in the code, in every report header and in the README — Etsy listing $0.20, transaction 6.5%, processing 3% + $0.25, offsite ads 12%; Gumroad 10% + $0.50. Every rate is overridable per run (--etsy-transaction-pct 7.0), because platforms change schedules and a tool that hard-codes "current fees" is a tool that's quietly wrong by next year.
The defaults' simplifications are written down, not buried. While checking the rates against the platforms' published pages (2026-11-04) I found two wrinkles the defaults simplify, and they ship in the README's Honest limits section, with the fix commands:
- Etsy's Offsite Ads fee is 15% for shops under $10,000/year and 12% above it. The default is 12%; smaller shops should run
--etsy-offsite-ads-pct 15. - Gumroad charges card processing (2.9% + $0.30) on top of its 10% + $0.50 on direct sales. The default models the platform fee only;
--gumroad-pct 12.9 --gumroad-fixed 0.80folds processing in.
Break-even assumes no returns, and says so in its output. For a returns-heavy listing the true break-even is higher, and the tool won't pretend otherwise.
Exit 1 is a finding, not a failure. A report with a loss-maker exits 1. Scripts and agents get the verdict in the exit code and in the --json output (loss_maker: true), which uses the same numbers as the text report — errors in JSON mode are {"error": ...} objects, argparse errors included, so an agent never has to parse prose to find out a run failed.
The boring parts that make it trustworthy
- 45 tests, standard library only. The sample's four listings are checked against numbers worked by hand from the stated rates — never copied from the tool's own output. Break-even prices round-trip: solve for 20%, feed the price back through the margin math, get 20% back.
- Validation names the row.
returns > units, a fractional unit count, an unknown platform, a non-numeric CSV cell — clean errors that say which row, never a traceback, never a silent zero. - Local only. It reads the file you point it at. No network code, nothing stored, nothing sent. Your margins are your business.
Try it
pipx install seller-profit-truth # or: pip install seller-profit-truth seller-profit-truth report examples/sample.json # from a clone seller-profit-truth breakeven --platform etsy \ --shipping-cost 6.25 --cogs 5.50 --target-margin 20 --units 30 Python 3.10+, zero dependencies, MIT. The repo ships a SKILL.md at the root, so AI coding agents can drive it directly — commands, JSON schemas, exit codes, and the rule that matters most: if you mention a listing's sales, mention its net and margin too.
Repo: https://github.com/jasveerai88/seller-profit-truth
If your dashboard and your bank account disagree, run the subtraction. The mug might not be the only one.