The rules we hold ourselves to.
These started as internal rules for what the scanner is allowed to say. They're published because a competitor running a hardcoded quiz can't make any of these claims, and because writing them down in public is the only version that costs us anything.
We never claim an absence.
The scan reads one page. A tag missing from that HTML may still be firing through a tag manager, and social links missing from a header may sit in a footer we could not render. So the report says "we could not find a tag on the page we read", never "you have no analytics". A component with no data scores nothing — never zero.
The score and the coverage are two different sentences.
You get a number out of 100 across what we could actually check, and separately how much of the full picture that represents. Blending them produces nonsense, and it flatters us: a site where only the easy checks ran would otherwise score suspiciously well. Under 60% coverage we withhold the letter grade entirely.
We never name a business we are critiquing.
We are a marketing agency, and the businesses we would be critiquing are the ones we want to sell to. Publishing damaging claims about a commercial rival has a name and a legal shape, and no amount of being right about the meta description is worth it. Findings are aggregated or anonymised. Named teardowns only where the business asked for one.
No number without a source.
If a figure appears anywhere on this site, it is traceable to something we can show you. This one is written down because we broke it: the site once claimed 2,847 businesses scored when the true number was one. It was removed, along with two invented testimonials quoting results we have never produced. It must not come back.
Nothing publishes under your name without a person reading it.
Drafting is automated. Publishing is not. We have seen a generator with a thin brief write a thousand words of confident, entirely invented pricing while every automated check passed it. A static gate cannot replace someone reading the thing.
Where we couldn’t see something, we left it unscored.
Why publish these at all
Because they're the part of the product that's hardest to copy. Anyone can build a scoring quiz in a weekend. What they can't easily do is hand back a report that admits what it couldn't see, because a number with holes in it is a harder thing to sell than a confident one.
And because a rule you've published is a rule you can be held to. If you find something on this site that breaks one of these, tell us and we'll fix it.