How Seryvon calculates scores and coverage

Seryvon · Published by Powehi · Updated 9 September 2026

Seryvon calculates website audit scores from collected signals and explicit rules. Its five pillars organise the findings; they are not scores supplied by search engines. This methodology describes the current public implementation, including the coverage adjustment that can lower a score when evidence is missing.

Collect evidence before interpreting it

An audit discovers URLs, crawls within its configured scope, extracts signals and adds available connector measurements. A rule evaluates these inputs and produces a result with evidence, weight and status. The calculation is deterministic for fixed signals and configuration. A new request to a website or provider may return different data, so a repeated live audit need not return the same score.

Calculate each pillar

First, exclude results marked not_applicable from the eligible set. Within that set, exclude not_measured from the measured average. Calculate coverage as measured criterion weight divided by eligible criterion weight. The current implementation multiplies the weighted measured average by that coverage, then rounds the pillar score to two decimal places.

coverage = measured weight / eligible weight
measured average = sum(score × weight) / measured weight
pillar score = measured average × coverage

For example, suppose three eligible criteria have equal weights. Two score 80 and 100; the third is unmeasured. The measured average is 90 and coverage is two thirds, approximately 0.6667 after the engine's coverage rounding. The displayed pillar score is 60.00. Missing data is not assigned a fabricated criterion value, but it still reduces this pillar's score through coverage.

Combine the five pillars

Default weights are SEO 30%, GEO 22%, GSO 18%, AEO 15% and ASO 15%. The global score is the weighted average of contributing pillar scores, renormalised if a pillar has no measured criterion. A completely unmeasured pillar does not contribute. If nothing contributes, the implementation returns zero; that must be interpreted alongside the absence of evidence.

Global coverage is calculated separately over distinct eligible criteria. Do not derive it by averaging pillar percentages: a criterion can belong to more than one pillar. Configuration can override weights and thresholds, so retain the actual measurement profile with any published result.

Read the labels correctly

Complete: coverage at least 90%.
Substantial: at least 70% and below 90%.
Partial: at least 40% and below 70%.
Insufficient: below 40%.

These are coverage tiers, not confidence intervals or success probabilities. Criterion statuses use separate thresholds: ok at 80 or above, warning from 50 to below 80, and critical below 50. Inspect the actual failed rule and evidence before choosing a remedy.

Compare compatible measurements

The comparison engine checks profile differences. Identical profiles are exact; allowed version differences may be compatible. Changed connectors or criterion overrides may require a comparison on their intersection. Changed rules, thresholds or pillar weights can make profiles incompatible. A descriptive side-by-side remains different from a valid improvement claim.

For a before-and-after review, keep the target scope, configuration and measurement sources consistent. Save the reports, inspect coverage changes and verify the deployed fix. Track business outcomes separately: technical improvement does not establish increased traffic or revenue.

Implementation and worked example

This explanation follows the scoring engine, default configuration and comparison rules. See the illustrative report, limitations and quickstart for practical use.