A good contribution is a fact you can point at
Short, specific, and written before the first bad submission rather than after it. Everything below is the evidence bar restated for a different situation — there is only one rule here, applied consistently.
What a good contribution is
A fact, and the page you read it on. Not an impression, not a recollection, not a summary of what you believe to be true. The test is whether a stranger can open your URL and find your sentence there.
Vendors are welcome, and get labelled
You may answer questions about your own product. Tick “I work on this” and the answer is published as yours — a vendor’s own statement is legitimate evidence and must be visibly theirs. What you cannot do is edit a price, a benchmark, or your own category; those go through the dispute route, with a source.
What gets rejected, with real examples
Three that come up constantly. In every case the problem is the gap between the claim and the source, not the opinion behind it.
“Each sandbox runs with network access disabled by default.”
Verbatim from the vendor’s own docs. Answers the question exactly. Published.
“Yes, it definitely supports network isolation — I use it daily.”
No source. Probably true, still unpublishable: nobody else can check it.
“It supports hybrid search.” — sourced to a blog post about a different product
The URL resolves and the sentence exists, but not on a page about this product. This is the most common near-miss.
Nothing is deleted when corrected
A superseded value keeps its date and stays on the page. If you read $3.35 last week, you can still see why. Deleting corrected data would make the archive a lie about what we published.
How moderation works, and how to appeal
One operator, checking your quoted sentence against your URL. That is a visit, not a judgement — which is the only reason a community this size is tractable at all. Every decision reaches you with the ground quoted and the moderator named, and you can appeal; the original text is kept, so if we were wrong it can still be published.
AI-assisted contributions
The defensible line is not “no AI” — that is unenforceable, and we use it ourselves in the ingest. It is the evidence bar, applied identically: a claim with a real source that says what it claims is fine however it was drafted. A plausible-sounding answer whose source does not support it is rejected. Repeated fabrication loses the account. That rule needs no detector.
No points, no levels, no badges
Accuracy — survived review ÷ submitted — is the only number on a profile. Invented scores make contribution feel like a game and answers feel like currency, and a directory whose whole claim is accuracy cannot also reward volume.
No account required
Nothing in the contribution flow needs one. When sign-in exists it belongs at submit, and even then anonymous-but-sourced stays possible — a sourced claim is checkable whoever made it.
- Sell a ranking position.
- Hide a price because a vendor dislikes it.
- Publish a number without a source.
- Delete a corrected value instead of superseding it.
Ready to contribute
Ready? Answer an open question, or dispute a value that is already published.