How is the Argos audit score calculated?

Four domains, one overall score

Argos groups scorable checks by business domain. Module and tool names indicate where measurements come from; they do not create additional categories.

When all four domains are measurable, each represents 25% of the overall score. If an entire domain cannot be measured, its share is distributed equally among the remaining measurable domains.

How points are awarded

1. Select applicable checks. Policy v12 assigns each known key a domain and a specific weight. Informational findings, unknown checks and unavailable measurements do not deduct points.
2. Calculate points. A PASS receives its full weight, WARN half and FAIL zero. Some graduated checks, including Lighthouse category scores, use their continuous ratio from 0 to 1 instead of this status rule. Displayed severity does not determine weight.
3. Normalize each domain. For a measurable domain, Argos divides earned points by the sum of weights of checks actually scored. The average of measurable domains then gives the overall score out of 100.

The report also shows measurement coverage: measured weights compared with expected catalogue weights, excluding checks that do not apply. Insufficient coverage makes the score provisional. If no domain is measurable, the score is unavailable.

Argos and Lighthouse measure different things

Argos gathers its own HTTP, HTML, browser and security findings. It also imports four Lighthouse category scores when that module responds: Performance, Accessibility, Best Practices and SEO. Each imported score becomes one weighted check in the corresponding Argos domain; the overall Argos score is therefore not the Lighthouse Performance score.

For example, the Lighthouse Performance score has weight 22 in the Argos Performance domain. A Lighthouse result of 80/100 contributes 17.6 of those 22 points, before other domain checks. Lighthouse computes its own Performance score from weighted metrics, according to its published Chrome for Developers method.

Worked example with an unavailable module

This deliberately reduced simulation uses real keys and weights from policy v12. It represents no website or client report. Suppose Lighthouse is unavailable: its checks contribute neither earned points nor each domain denominator.

Overall score: 74/100 = (72.7 + 44.4 + 78.6 + 100) ÷ 4, rounded to an integer.

Unavailable Lighthouse is not treated as a failure. All four domains still have measured checks here, so each retains 25% of the overall score. Coverage signals the many missing measurements and makes this score provisional. Other unavailable modules or additional checks would change denominators and the score.

What one score cannot guarantee

• The result describes one URL and one moment: content, network, browser and external services can change measurements.
• Scores are most comparable when policy version, fingerprint and coverage are consistent. Reports preserve this information to explain differences.
• A high score does not certify security, accessibility or search visibility for an entire website. Human checks and untested journeys remain necessary.

For the checks discussed here, see the official references: OWASP security headers and W3C WCAG overview. The Argos score is not an assessment of conformance to those frameworks.