Transparency
How the score is calculated
Every SEO tool gives you a health score and none of them will tell you the formula. Here is ours, in full, including the parts that are a judgement call rather than a measurement — and an explicit list of what this inspection cannot tell you.
- Standard
- SW-1:2026.2
- Effective
- 2026-07-01
- Recheck window
- 28 days
- Certificate validity
- 180 days
1. The scoring formula
The score is a weighted percentage. Each of the 45 checks carries a fixed weight. A check returns one of three results, which map to a fraction of that weight:
| Result | Value | Meaning |
|---|---|---|
| Pass | 1.0 × weight | Meets the published pass criterion in full. |
| Advisory | 0.5 × weight | Meets the advisory criterion. Recorded, half credit. |
| Fail | 0.0 × weight | Meets the fail criterion. No credit. |
| N/A | excluded | Not applicable (e.g. hreflang on a single-locale site). Removed from the denominator entirely — it neither helps nor hurts. |
Formula
score = round( 100 × Σ(result_value × check_weight) / Σ(applicable_check_weight) )Category scores use the same formula restricted to that category's checks. Because N/A checks are removed from the denominator, a site is never penalised for lacking a feature it has no reason to have.
Worked example
A site passes every check in Crawlability except CR-03 (redirect chains, weight 2) which returns advisory, and CR-06 (crawl budget, weight 2) which fails. Category weight available is 16.
earned = 16 − (2 × 0.5) − (2 × 1.0) = 13
category score = round(100 × 13 / 16) = 81You can recompute any score on any report we issue by hand, from the published weights. If our arithmetic disagrees with yours, that is a bug and we want to hear about it.
2. Pass or fail is not the same as the score
This is the most important design decision in the standard, and it is borrowed directly from vehicle inspection. A car with four excellent tyres, perfect lights and no brakes does not get a warrant with a note about the brakes. It fails. Certain faults are disqualifying regardless of how good everything else is.
Immediate fail. Certificate cannot be issued at any score.
3 checksFails the item. Two or more major fails is an automatic overall fail.
10 checksFails the item and deducts weight, but does not block a pass on its own.
24 checksRecorded on the checksheet. Does not affect the determination.
8 checksDetermination logic
if (any critical fault) → FAIL
else if (major faults >= 2) → FAIL
else if (score < 60) → FAIL
else if (score < 75) → CONDITIONAL
else if (major faults == 1) → CONDITIONAL
else → PASSA conditional result means the site is close and the remaining faults are scoped and known. It is not a pass and no certificate is issued, but it is a materially different conversation from a fail.
Grade bands
90–100
Fit
75–89
Fit with advisories
60–74
Conditional
40–59
Fail
0–39
Serious fail
The grade describes the score. The determination describes fitness. A site can grade B and still fail on a single critical fault — and the report always states which of the two is the reason.
3. Where the weights come from
Category weights total 100 and are set on one principle: how completely does a fault in this area prevent the site from competing at all? Crawlability and indexation carry the most weight because a fault in either makes every other category irrelevant — perfect content on a page that cannot be crawled scores zero in the only place that counts.
We want to be direct about this: these weights are a considered judgement, not a measurement. Nobody outside a search engine knows the real coefficients. What we can promise is that the judgement is published, applied identically to every site, versioned, and revised on the record.
4. Every number carries its provenance
A lab performance figure and a field performance figure are not the same kind of fact, and presenting them identically is how tools quietly mislead people. Every value on a SEOWof checksheet is tagged with one of three confidence classes.
Directly observed by the inspection — a response code, a header, a rendered DOM, or field data from real users. The highest-confidence class.
Computed or inferred from observed data — a lab performance estimate, a content comparison against a ranking set, or an indexation estimate from impression differentials. Sound, but an inference.
Supplied by you — your Business Profile details, your priority pages, your target queries. We validate consistency but we cannot verify the underlying truth.
5. What this inspection cannot tell you
Most tools imply total coverage by omission. We would rather state the boundary, because a limit you know about is manageable and a limit you do not know about is a liability.
Content quality in an editorial sense
We can measure structure, depth relative to the ranking set, and uniqueness. We cannot measure whether your writing is any good, whether your argument is correct, or whether an expert would respect it. No automated system can, and any tool claiming a "content quality score" is measuring proxies and calling them judgement.
Brand strength and demand
Branded search volume, reputation and the trust a buyer places in your name are the largest determinants of commercial search performance and sit almost entirely outside a technical inspection. A site can pass every check and still lose to a better-known competitor.
Whether your strategy is right
We check whether a page matches the intent of the query it targets. We cannot tell you whether that query is worth targeting for your business, whether your margins support the acquisition cost, or whether search is even the right channel.
Competitor internals
We can compare your public signals against competitors' public signals. We cannot see their analytics, their conversion rates, or what they are spending. Any competitive claim we make is bounded by what is externally observable.
How a search engine actually weights anything
No third party has access to a ranking system. Our weights are a considered, published judgement informed by documented behaviour and observed correlation — they are not a reverse-engineered formula, and we do not present them as one.
Future performance
An inspection is a point-in-time assessment of fitness against a standard. It is not a forecast. We do not publish traffic projections because we cannot make honest ones.
6. How often to re-inspect
A certificate is valid for 180 days. That number is not arbitrary — it reflects how quickly a live website drifts out of compliance. In our experience the large majority of serious faults are introduced by the site's own deployments, not by anything a competitor did.
| Site profile | Cadence | Reason |
|---|---|---|
| Brochure site, rarely changed | 6 months | Drift comes from platform and plugin updates rather than your own changes. |
| Content site publishing weekly | Quarterly | Index bloat, cannibalisation and orphaning accumulate with publishing volume. |
| E-commerce with changing catalogue | Monthly | Discontinued products create 404s and lost equity continuously. |
| Site shipping code weekly | Continuous | Any deploy can introduce a critical fault. Monitoring catches it in days, not quarters. |
| After any migration or replatform | Immediately | Migrations are where the most severe faults are introduced, and where equity is lost permanently if not caught early. |
7. Versioning and revisions
The standard is versioned so a result is always reproducible against the criteria in force on the day. Reports record the version used and remain valid under it. We revise twice yearly, and every change is logged with its reasoning — including the ones where we got something wrong.
- SW-1:2026.2Current. Added the AI & Answer Engine Readiness category at weight 5, deliberately conservative while the evidence base matures. Reduced Authority from 9 to 7 in consequence. Downgraded meta descriptions (OP-02) from minor fail to advisory — they are not a ranking factor and scoring them as an error was overstating them.
- SW-1:2026.1Initial publication. 42 checks across 8 categories.