Methodology · Scoring
How the score is calculated. All of it.
Most tools treat their health score as a trade secret, which means you are asked to accept a number you cannot check. Here is ours in full — every weight, every step of the arithmetic, and what each of the 46 findings costs. If you cannot reproduce your own score from this page, that is a bug and we want to hear about it.
1
Every category starts at 100
There are 7 of them. A site with nothing wrong scores 100 in each.
SEOTechnicalPerformanceContentLocalSecurityAccessibility
2
Each finding subtracts its severity
Findings come off the category they belong to. Nothing is scaled, curved or weighted by how many pages are affected — one critical finding costs the same whether it is on one page or all of them, because one is already enough to be worth fixing first.
| Severity | Points off its category |
|---|---|
| Critical | −25 |
| High | −12 |
| Medium | −6 |
| Low | −3 |
| Suggestion | 0 — noted, never deducted |
A category cannot fall below 0 or rise above 100, so a site with six critical findings in one category scores 0 there rather than a negative number.
3
Two categories are capped by Google's own measurement
For performance and accessibility we take the lower of our score and Google Lighthouse's. If Lighthouse says your performance is 46, your performance score is 46 no matter how clean your markup is.
It only ever moves a score down. We would rather be judged on Google's measurement than on our own opinion of it — and if Lighthouse does not answer, the check is reported as not tested rather than quietly passed.
4
The categories are blended into one overall score
Two mixes. Which one you get depends on whether the local checks applied to your site.
| Category | Most sites | Local businesses |
|---|---|---|
| SEO | 25% | 22% |
| Technical | 25% | 22% |
| Performance | 20% | 18% |
| Content | 10% | 8% |
| Local | not counted | 12% |
| Security | 10% | 9% |
| Accessibility | 10% | 9% |
Local is left out of the overall for sites the local checks do not apply to, rather than counted as a perfect 100. A software company would otherwise be rewarded for opening hours it has no reason to publish. For a local business it takes its share from SEO, technical, performance and content, because for a shop those findings are not a footnote.
Worked example
Five findings, done by hand.
The totals below are produced by the same code that scores your report, so the arithmetic on this page cannot drift from the arithmetic in the product.
SEO
100 − 25 Homepage is set to noindex − 6 Meta description missing = 69
Technical
100 − 12 No mobile viewport tag = 88
Performance
100 = 100
Content
100 − 6 Two thin pages = 94
Security
100 − 3 Referrer-Policy not set = 97
Accessibility
100 = 100
Overall
(69 × 25%) + (88 × 25%) + (100 × 20%) + (94 × 10%) + (97 × 10%) + (100 × 10%) = 88
Rounded to the nearest whole number at the last step, and only at the last step.
Every finding
What each of the 46 checks costs.
The complete list. Nothing is held back, and every one links to the page explaining how to fix it.
SEO
14 checksTechnical
6 checksPerformance
4 checksContent
1 checksLocal
10 checksSecurity
7 checksAccessibility
4 checksWhy publish this
A score you cannot check is a score you cannot argue with, which is convenient for whoever is selling it. We would rather you were able to disagree with us: if a finding is wrong for your site, you can see exactly what it cost you and decide for yourself whether we were right.
It also keeps us honest in a way a promise would not. Our own site takes this same audit every day and we publish the result, pass or fail — and the release rule is that a build does not go live below 90 with any critical finding open.
See today's score for this site →
What we check, and what we cannot →
Check our arithmetic
Run it on your site and add up the deductions.
Free, no account. Every finding in the report names its severity, so you can work the total yourself.