The standard
The Evidence Grade is the four-tier standard we apply to every claim in the library — so you can see at a glance how much of your roadmap it can carry.
Why it exists
A statistic gets published once, quoted without its method, and repeated until it feels true. By the time it reaches a roadmap nobody remembers whether it came from a peer-reviewed trial of 40,000 people or a vendor’s landing page.
Product evidence means the grade travels with the claim. The number and its weight arrive together, so a decision can be argued on the strength of what is behind it rather than on who said it most confidently.
We found the original paper, report or dataset, read it, and confirmed the number says what the claim says it says. Author, year, venue, sample size and effect size are all recorded. Replication status is checked where the field has one.
What we do with it: Safe to build on. We will say it on camera.
The source exists and the finding is genuine, but the methodology, the interpretation or the generalisation is actively argued over — or the effect shrinks badly outside the original setting.
What we do with it: Usable with the caveat stated out loud. Never quoted as settled.
The claim traces to a blog citing a blog, a vendor case study with no method, an unpublished deck, or a study so small or so old that it cannot carry the weight being put on it.
What we do with it: Reported as folklore, not fact. We say where it came from and why it is thin.
The claim died under a primary-source check: the study does not exist, the number was invented downstream, the original said something materially different, or the cited organisation denies it.
What we do with it: Removed from the library and added to the topic blacklist, with the source that killed it.
How a claim gets graded
Not the article that quoted it. The paper, report or dataset itself. A claim whose original cannot be located cannot rise above SHAKY.
Sample size, population, effect size, and what was actually measured. A great many famous numbers describe something narrower than the sentence they now appear in.
Has it survived scrutiny, failed to replicate, or been walked back by its own author? Real findings that are argued over become CONTESTED, not SOLID.
Author, year, venue, sample, effect and link, written down so anyone can repeat the check. Every graded claim in the library carries this trail.
Use it yourself
A standard is only worth something if other people apply it. Grade your own roadmap claims with it, cite it in a document, or hand it to whoever is arguing that a competitor’s pattern must work because a competitor is using it.
Cite it as The Evidence Grade — LAUNCHED, and link to this page.