Designing deterministic website diagnostics

Begin with a bounded question

Website audits become vague when they attempt to grade quality in the abstract. A deterministic diagnostic instead defines a finite question: can this public page be found, parsed, understood and trusted under an explicit set of checks?

Each check needs a source input, a pass condition, a point value and a remediation directive. If any of those are missing, the score is difficult to reproduce or challenge.

Capture evidence before calculating a score

Extraction and evaluation are separate responsibilities. The extractor records status, headers, HTML, metadata and machine files. The evaluator receives that normalised evidence and applies pure rules.

This separation makes failures inspectable. It also allows a fixture to reproduce the evaluation without repeatedly requesting the live site.

• Input evidence is retained
• Rules are explicit
• Identical evidence produces an identical score
• Unavailable evidence never silently passes

A score is a summary, not the product

A single number is useful for orientation, but it cannot be the only output. The diagnostic must show which evaluations passed, which failed, how many points remain recoverable and what change would alter the result.

Percentiles and visibility-growth claims require a real benchmark dataset. Without one, deterministic score bands and remediable points are the more honest language.

Verification closes the loop

A remediation directive is complete only when the system can be scanned again and the changed evidence produces the expected state. The method therefore supports implementation, regression testing and dispute resolution—not merely lead generation.