Skip to content
Saufee

The fee dataset

Every rate behind this site, as open data. 16 platforms, each fee carrying the link it was read from and the day it was read. Free to use, including commercially, with attribution.

Why this exists

Fee figures circulate the internet with no date and no source on them. An article says Etsy takes 6.5%, a forum reply says 5%, a calculator says something else again, and none of the three says when it was checked or where it came from — so a seller cannot tell which is stale. Every number here carries both, and the whole set is downloadable so you can check it yourself rather than take our word for it.

What is in it

For each platform: the itemised fee lines, the formula behind each one, the settings that change the answer, the source links, the date read, and an explicit list of what the model does not capture. The fee lines are generated by running the same engine that powers the calculators on this site, at five reference sale values from $10 to $2,000 — so the dataset cannot quietly disagree with the pages, and the spread makes percentage and flat components obvious.

Those reference values are not round numbers for the sake of it. $10 sits exactly on eBay’s per-order fee step, which is the sort of threshold a single worked example hides.

What it does not cover

  • Seller-side fees only. Buyer-paid fees are never deducted from a payout.
  • United States rates on default settings. Many of these schedules differ by country.
  • Sales tax and VAT are excluded: they are collected from the buyer and passed on, so they are not a cost to the seller.
  • Negotiated or volume pricing is not represented. Only published standard rates can be sourced.
  • Chargebacks, disputes and refunds are real costs but are not part of the price of a completed sale.
  • Each platform carries its own notModelled list. Read it before relying on that platform.

How it is kept current

An automated check runs every Monday and re-fetches all 16 source pages, confirming the text each rate was read from is still there. It never edits a rate: scraping a number and publishing it unread is exactly the guesswork this dataset exists to replace. It raises a flag, and a person re-reads the schedule by hand. The full method, including what that check does not prove, is on how we verify.

Licence and citation

Published under CC BY 4.0. The rates themselves are facts and nobody owns them; what is being given away is the collection, the sourcing and the dating. Attribution is the only condition.

What each platform contributes

PlatformCategoryFee linesSourcesGaps declaredRead
StripePayments2122026-08-28
PayPalPayments2122026-08-28
SquarePayments2122026-08-28
EtsyMarketplace3232026-08-28
eBayMarketplace2132026-08-28
AmazonMarketplace1152026-08-29
MercariMarketplace1142026-08-29
ShopifyStore2132026-08-28
GumroadDigital goods2112026-08-28
Lemon SqueezyDigital goods2112026-08-28
VintedMarketplace1122026-09-01
DepopMarketplace1122026-09-01
FiverrMarketplace1132026-09-01
PoshmarkMarketplace1152026-09-05
StockXMarketplace2152026-09-05
UpworkMarketplace1142026-09-03

“Gaps declared” counts what each platform’s entry explicitly says it does not model. A high number is a good sign: it is the difference between a dataset that knows its limits and one that pretends it has none.