How Scores Are Computed

Every score in HubSEO has a published formula, hard bounds, and an explicit missing-data rule. This page is the contract. If a report and this page ever disagree, treat that as a bug.

Overall audit score: seo-weighted-v2

overall = round(0.45 * performance + 0.35 * seo + 0.20 * bestPractices)

• Inputs: mobile Lighthouse Performance, SEO, and Best Practices category scores.
• Accessibility is reported separately, with weight 0.
• Bounds: 0..100.
• Missing-data rule: if the mobile lab run is unavailable, the audit is partial and the overall score is omitted. No missing category is treated as zero.
• This is a mobile lab diagnostic. It is not a ranking prediction.
• Explaining a change: compare the three category scores first. Each mapped category retains its Lighthouse audit refs, which identify which findings moved. Lighthouse's internal audit weighting inside a category is not re-implemented, so a category delta is the most precise arithmetic attribution available.

Check-set score

score = round(100 * (passes + 0.5 * warnings) / count(non-info checks))

• info checks are unavailable evidence: excluded from the denominator, not counted as passes or failures.
• Bounds: 0..100. A list with only info checks returns null.
• Explaining a change: diff the non-info check IDs and statuses. One pass flip is worth round(100 / n) where n is the current measured count; a warning is half credit.

AI readiness score

score = round(100 * count(pass) / count(measured signals))

• Measured: pass, fail, warning. unavailable is excluded.
• A warning is measured but earns no partial credit.
• Bounds: 0..100, null when nothing was measured.
• Grades: Strong ≥ 75, Mixed ≥ 50, Limited < 50, unmeasured when null.
• Full detail: AI readiness.

CrUX Core Web Vitals grades

• Real-user p75 values from the Chrome UX Report, graded per metric against the standard good / needs-improvement / poor thresholds.
• A missing metric is ungraded and omitted. It is never converted to zero.
• Lab and field stay separate: Lighthouse lab timing never masquerades as real-user INP. See Core Web Vitals and CrUX.

Lighthouse category scores

• Native Lighthouse category scores, mapped to 0..100.
• Unavailable when a Lighthouse run fails; reported as unavailable, not zero.

GEO visibility aggregate

score = round(50 * presenceShare + 50 * citationShare)

• Requires at least 3 completed external samples. Below that, the status is insufficient_data, not a low score.
• Explaining a change: compare the cited and mentioned booleans of the completed samples.

PDF presentation colours

Presentation only: GOOD ≥ 90, WARN ≥ 50, BAD < 50; a null score renders as "Unscored" with muted colour. The colour never changes the underlying value.

Human severity

Belongs to a single finding, ranked by the rule in the Evidence Desk: Fix this first, Important improvement, Worth improving, Optional polish.

The standing rule

Scores do not predict traffic, rankings, revenue, or real-engine visibility. They describe the measured state of the URL at scan time. When evidence is missing, HubSEO reports the absence instead of computing around it.