Method
How we check a figure
Most of the work behind a page on this site is not writing. It is establishing which of four published values for the same fee is the current one, and who has an interest in the answer.
Last updated 8 August 2026.
The four checks every figure goes through
1. Where did the number originate
Trace it back until you reach a platform announcement, a help page, a regulation or a government publication. A figure in an article that cites another article is not sourced, it is repeated. Most marketplace fee content is at least two hops from a primary source and never says so.
2. What date does it belong to
Fee content ages badly and is rarely revisited. During research for the reimbursement guide on this site, three different effective dates for the same policy change were found in circulation, one of which corresponds to nothing the platform published. A rate is a statement about a moment; the moment gets recorded alongside it.
3. Who benefits from the number being what it is
This is the check that changes conclusions most often. If every published estimate of a loss comes from companies that sell recovery software, that is not a consensus, it is a marketing category. The publisher of each secondary source is named in our source lists so the reader can apply the same discount we did.
4. Do the sources agree
When secondary sources disagree about a figure the platform publishes, we go to the platform and print its number; the disagreement is noted only to explain which circulating figures are stale. When the platform publishes no figure — Buy Box tolerances, advertising benchmarks, buyer behavior — the estimates are shown side by side, each labeled with its publisher. Until September 2026 this site treated the spread itself as the finding; the audit that month showed that practice let wrong figures survive as long as a second source repeated them, and the standard was changed.
What that produces on a page
| Element | What it guarantees |
|---|---|
| Sources block with retrieval dates | Every claim can be checked at its origin, and you can see how old the check is |
| Publisher named for secondary sources | You can see who had a commercial interest in the figure |
| Contested-figures table | Where sources conflict, you get both values instead of one confident guess |
| Illustrative figures labeled | You can tell a demonstration of arithmetic from a quoted rate |
| Published and modified dates | You can judge whether a page has been revisited since the last fee change |
Software reviews
No software reviews are published on this site yet. The rubric below is published in advance so that the standard exists before the first review does, rather than being written to fit a conclusion afterwards.
| Criterion | Weight | What earns a high score |
|---|---|---|
| Data accuracy | 30% | Figures reconcile against the platform's own reports on a sampled basis, with the sample stated |
| Operational fit | 25% | Solves a problem a working seller actually has, without a workaround |
| Cost at real volume | 20% | Total cost stays predictable as order volume grows |
| Setup and support | 15% | Time to first useful output; support answers operational questions |
| Data portability | 10% | Exports are complete and usable outside the tool |
Conditions that will apply
- Tools are tested on paid plans bought at list price. Vendor-supplied review accounts are declined.
- Every review states the plan tier tested and the month of testing.
- No score is assigned to a tool that has not been run on real data.
- Scores are revisited when a vendor changes pricing or ships a material change.
- No sponsored reviews, no paid placement, no vendor sight of a review before publication.
- This site carries no affiliate or referral relationships with any reviewed tool. Funding is advertising only, and it is never a scoring input.
Corrections
Errors are corrected in place with a dated note and logged on the corrections page. Nothing is edited silently.