Skip to main content

Methodology

A score you can explain and reproduce.

From the server response and rendered page to scores, findings and priorities. See what was measured, what was missing and how to read the report.

Methodology v1.4, Updated: 24 Oct 2026, Changelog

Source before score

Every finding shows where its signal came from.

Thresholds from standards

Every threshold has a source: Google documentation, WCAG or OWASP.

A repeatable measurement

Scope, device and method version are saved in the report.

Score

From signal to score.

Each module uses completed checks. Weight reflects the impact of a problem, while missing data stays visible as N/A.

  1. Source data Lab measurement, server response, rendered DOM, crawl
  2. Threshold Compared against a public standard or reference configuration
  3. Finding pass, warn, fail, or N/A when data is missing
  4. Severity & confidence Finding impact × automated detection confidence
  5. Priority A shared P1–P4 scale across all modules
  6. Score 0–100 Weighted aggregation of a module’s checks

Example finding

How to read one finding

Using a missing HSTS header: where the signal came from, which rule applied and how it changed the module score.

Every finding in the report carries the same fields.

Security company.com example
Missing HSTS header P1, fix first
Signal
missing Strict-Transport-Security header
Source
HTTP response
Rule
OWASP Secure Headers
Scope
entire domain
Confidence
high
Severity
high
Module impact
fail, high weight, 0 pts from this check
Score aggregation rules

A module score comes from available, completed checks. N/A checks are neither a pass nor a zero and do not enter the denominator.

Module score

Points from completed checks ÷ sum of those check weights × 100. Fail = 0 for that check, warn = partial weight, pass = full weight. Performance is the exception: since 1 Oct 2026 its score is the Lighthouse score from the lab run.

N/A outside the denominator

A missing signal does not lower the score and is not counted as a pass. We show N/A separately so “not measured” is never confused with “zero”.

Overall score

Combines measured module scores using the weights of the published methodology version, then applies the risk gate. An N/A module does not enter that aggregation.

Risk gate

One very weak module limits the whole score. When the weakest measured module scores below 50, the overall score is capped at 69; below 40, it is capped at 59. An N/A module does not trigger the gate. A module whose measurement is flagged as unreliable (for example a Lighthouse run that stayed anomalous after a re-measure) keeps its weight in the average but cannot cap the score.

Priority ≠ score

P1–P4 orders the implementation backlog. It does not map 1:1 to score points, because severity, scope and confidence set priority separately.

Weights and aggregation

Checks do not count equally: the weight depends on impact (high > medium > low). A passed check earns its full weight, a warning part of the points, a failure zero. A module score is earned points over maximum points, scaled to 0–100.

Missing data (N/A)

A module or check without data is reported as N/A and excluded from the denominator. The score covers only the measured scope and coverage. A deeper scan level adds modules and scope, but does not rewrite verdicts of checks already measured.

Data

Start with the source.

Lab measurements, HTTP responses, automated detection and CO₂e estimates have different limits. The report labels each one.

Lab measurement and field data are two different sources.

The Performance module score is the Lighthouse score from a lab measurement. CrUX field data sits next to it, with its own source.

Lab measurement

  • Lighthouse loads the page in a controlled environment at scan time.
  • It shows what slows the page down and where to start fixing.
  • It is a single run, so the score can move a few points between runs.

Field data (CrUX)

  • The Chrome UX Report describes visits by real Chrome users over the last 28 days.
  • We report LCP, INP and CLS at the 75th percentile, which covers most visitors.
  • When a site has too little traffic, the field shows N/A with the reason.
Lab measurement

Lab performance metrics

Lighthouse / PageSpeed API

Source limit: Results may vary between runs

Direct measurement

HTTP checks

Statuses, headers, redirects, cookies

Source limit: Server state at scan time

Automated detection

WCAG tests

Automated, on the rendered DOM

Source limit: Some criteria require manual review

Model estimate

CO₂e estimate

The SWDM v4 model based on data transfer

Source limit: Not an emissions measurement or a certificate

First-party data

Crawl

Links, sitemap.xml and a manual list, the page is rendered with JS

Source limit: Page limit per plan; recurring issues are grouped into one finding with its scope

Two analysis inputs

Insight does not install a script on the site. It reads what the browser shows and what the server returns, in one scan session.

Rendered page

A real browser, JS, DOM and what the user actually sees.

Server response

Statuses, headers, redirects and cookies at scan time.

No script on the client site

Thresholds

Thresholds have sources.

Checks use thresholds from source standards. You can cite those standards in your client report.

The Insight 0–100 scale

A score takes its colour from the band it falls into, and nothing else. The report and the app use the same bands.

  • 95
    ≥ 90 good standing sample score: 95
  • 80
    70–89 needs work sample score: 80
  • 60
    50–69 significant issues sample score: 60
  • 35
    < 50 high priority sample score: 35

The overall score is a weighted average of the measured modules on a shared 0–100 scale. Only the Performance module takes the Lighthouse score as is. No score is a compliance certificate or a legal opinion.

Thresholds from source standards

Core Web Vitals

LCP
≤ 2.5 s
INP
≤ 200 ms
CLS
≤ 0.1
Chrome UX / web.dev

WCAG 2.2 (AA)

Text contrast
≥ 4.5:1
W3C

Security headers

CSP
present
HSTS
present
X-Content-Type-Options
nosniff
OWASP

Priorities

A score is not a repair plan.

P1–P4 is implementation order, not a synonym for severity. Severity comes from the threshold; priority combines severity, scope and confidence. P1 is not always a critical defect. Sometimes it is a wide, high-confidence issue. Estimated effort is only an auxiliary recommendation.

What sets the order

  • SeverityComes from the threshold, so the same problem gets the same severity.
  • ScopeAffected page count moves a finding along the P1–P4 scale.
  • ConfidenceAutomated detection confidence affects position on the scale.
  • Estimated effortAn auxiliary recommendation. It does not lower severity.

P1–P4 scale

  • P1Fix first
  • P2Plan it
  • P3Fix when convenient
  • P4Backlog / low

N/A: not enough data. Never zero or success.

Another example: text contrast

Contrast below 4.5:1, P2

Signal
text below WCAG AA 4.5:1
Source
rendered DOM
Rule
WCAG 2.2, 1.4.3 Contrast (Minimum)
Scope
2 pages
Confidence
high
Severity
medium
Module impact
fail, medium weight, 0 pts from this check

Scan levels

Four levels, different scope.

A deeper level adds scope, but it does not change verdicts of checks already measured.

Quick

Quick screening of a single URL. Included in Free

No Lighthouse

Modules5
Time~1 min
Cost4 cr

Standard

Regular site quality control. Once a month on Free

Adds performance

Modules10
Time3–5 min
Cost12 cr

Complete

Audit before a release or acceptance. Growth and up

Security cluster (ZAP)

Modules22
Time8–10 min
Cost18 cr

Comprehensive

Competitor comparison and the widest scope. Growth and up

Competitors + AI Search readiness

Modules31
Time12–15 min
Cost25 cr
See costs on the pricing page
Understanding scan scope and cost

Each level runs a defined module set. Credit costs match the pricing page.

Module count = check suites from the published module catalog. Some suites report into a shared module (e.g. accessibility and privacy screening), so a report may show fewer sections than suites. Basic security checks (TLS, headers, cookies) run from Standard; the OWASP ZAP vulnerability cluster is added at Complete.

AI

AI explains. It does not measure.

A generative model does not set the raw measurement result and does not replace source thresholds. It may help group findings and draft recommendation copy.

Measurement layer

No AI

Measurement, thresholds and severity always come from source data.

pass, warn, fail, N/D

Description layer

With AI

Descriptions, grouping and recommendations, labelled, for review.

Explains the result, for review

Rules for AI use in reports

Measurement without a model

Scores, thresholds and severity never come from a generative model.

AI on raw scan data

It groups findings, describes impact and drafts recommendations.

Review only

AI recommendations are labelled and nothing is applied automatically.

Complete without AI too

The score and findings list do not depend on the generative layer.

Limits

What needs a separate review.

An automated scan detects some problems. These boundaries explain where its coverage ends and further review is needed.

WCAG / EAA

Does not confirm compliance

Replaces neither manual review nor legal opinion.

CO₂e

Not a measurement or certificate

An SWDM model estimate, not a lab result.

ZAP

Not a pentest

An automated vulnerability scan, not a manual penetration test.

Crawl

Not a full enterprise crawl

No guarantee of finding every URL on the site.

Lab vs CrUX

Not the same source

Shown separately, without treating lab data as field data.

See the technical crawl methodology

Comparability

Compare like with like.

Page content, third-party services and network conditions can change between scans. The report records the analysis conditions. Results are compared using the same scale and thresholds.

Time and version

Scan date and methodology version

Scope

Scan level, URLs and device profile

Completeness

Modules completed and marked N/A

Method changes

Communicated, never a silent backfill

A before and after comparison in practice: the cometweb.io case study

Sources

Standards and sources

Documentation for the thresholds and models used in the report. Refer to it when you need to check a rule or cite a source.

Changelog

Applies to reports from: 18 Oct 2026

v1.4 18 Oct 2026

Two model changes that change scores. Since 1 Oct 2026 the Performance module score is the Lighthouse score, with no extra deductions for optimisation suggestions. Since 18 Oct 2026 the Security score uses a new model, and older history points are marked so the trend does not show a jump that never happened. We added this entry late, on 24 Oct 2026.

v1.3 3 Oct 2026

Risk gate described in the aggregation rules: the 50 and 40 point thresholds and the exception for an unreliable measurement. The rule was already part of the calculation, so the description changes, not the scores.

v1.2 22 Jul 2026

Table of contents, aggregation and N/A rules, full signal→score example, explicit AI role, versioned sources.

v1.0 15 Jul 2026

First public methodology: scoring path, sources, thresholds, priorities, scan levels and limits.

See the methodology in practice.

The sample report shows the data sources, timing and limits of the analysis. The Free plan then lets you scan one project, no credit card required.

No credit card required Data sources in every report Thresholds from public standards