How the Website Score works

The principle: measured, or absent

Every input to the score is measured directly from your pages at the moment you run the check — the initial HTML, fetched the way search engines and AI crawlers fetch it. Nothing is estimated from industry averages, inferred from your sector, or invented to fill a gap. Where a check cannot run, it is excluded from the calculation entirely rather than scored as zero: a score never drops because data was unavailable.

The seven components

Thirty-plus individual checks are grouped into seven components. Each check carries a weight based on its severity — how much it typically affects whether customers find, trust and choose a business: critical checks weigh 8, high 5, medium 3, low 1. A component's score is the weighted share of its checks that pass, from 0 to 100. Checks reported as "not detected" (softer signals, where absence proves less) count at half weight against the score.

The overall Website Score is the weighted mean of the component scores using the weights above. The weights reflect our judgement, formed across real audits, of how strongly each area affects commercial outcomes for UK local businesses — they are opinions held consistently, reviewed with each methodology version, and published here so you can disagree with them intelligently.

Score bands

80 and above reads as Strong: the area is doing its job. 55–79 is Needs work: functioning, but leaving visibility or conversions unclaimed. Below 55 is Weak: the area is actively costing the business visitors or enquiries.

Confidence

Every report states its confidence. High means multiple pages were read with no fetch failures; medium means a single page was analysed; low means some pages could not be fetched, so treat the result as indicative. The report also states when it was measured — results are a snapshot, not a subscription.

What the score deliberately does not claim

The Website Score measures your website. It cannot see your Google Maps rankings, your reviews, your citations across directories, or who links to you — each needs a different data source, and every report lists these exclusions explicitly. A full picture of local visibility combines all of them, which is what our paid audit work does. We would rather show you a smaller, true number than a bigger, invented one.

Known limitations

Checks run on the initial HTML without executing JavaScript — deliberately, because that is how many AI crawlers read the web, but it means single-page-app websites can score lower on content checks than what a human sees. Detection is text-pattern based and errs generous: a false "missing" is treated as the worst error the tool can make. And a handful of checks (opening hours, credentials, reviews mentioned) detect textual signals, not verified facts.

Versioning

This is methodology version 1.0 (August 2026). Every report carries its methodology version, so scores from different versions are never silently compared. When the methodology changes, this page changes with it.