Methodology
How every number on this platform is produced — what is measured, what is inferred, and what is refused. If a figure cannot be traced to a source, it does not appear.
What an audit collects
An audit crawls your public site (structure, metadata, headings, links, word counts, structured data), loads key pages in a real browser to measure performance (LCP, CLS, TBT, transfer weight) and accessibility (axe-core, WCAG 2.1 A/AA), and reads every connected data source: Google Search Console, Google Analytics 4, PageSpeed Insights, the SERP provider, your repository and public app-store listings. Every measurement is stored as evidence with its source, collection date and reliability. Where a source is not connected, its data is reported as unavailable — never estimated.
The five evidence classes
Every finding carries the classification its evidence actually supports. Verified means directly observed or measured. Supported inference means the evidence supports the claim but causality is unproven. Hypothesis is worth testing, not a factual claim — hypotheses never affect scores. Data unavailable states honestly that no source provides the measurement. Conflicting evidence means two real measurements disagree (for example lab performance vs PageSpeed’s throttled mobile) and a person should decide which condition matters for them.
How scores work
Department health = 100 − Σ(severity penalty × count), capped at 0: critical −22, high −12, medium −6, low −2.5, info 0. Findings classified data-unavailable are excluded, because absence of data is not a defect. A department with nothing measurable shows a dash — never zero. Next to each score, a confidence label (high / medium / low) says how much data stood behind it: sources connected and findings scored. A 100 with low confidence means “nothing wrong was found in the little that was measurable” — not a clean bill of health. Opportunity priority uses the published formula shown under every score, weighted so relevance and impact beat raw volume.
Keywords: ranked, optimisation, new
Already ranking merges the SERP provider’s index (position, page, estimated volume) with your real Search Console queries (clicks, impressions, average position), each row naming its source. Optimisation opportunities are keywords already ranking at positions 8–30 — page one is within reach by improving the existing page. New opportunities are provider demand for topics you do not rank on: candidates must share a specific topic word with your proven keywords, pages or titles (generic words like “tool” or “free” are never enough), are deduplicated, mapped to an existing or proposed page, and each names the seed that produced it. Volumes are provider estimates for your configured market and are always labelled as such.
Experiments
An experiment is the measurement of one change. Press Test this on an opportunity: the current value of its metric is recorded as the baseline. You make the change on your site, audits re-measure the metric over the stated window, and the outcome — positive, negative or inconclusive — is recorded against the baseline in the experiment ledger. Outcomes are never projected; an experiment that did not move the number says so.
What each integration provides
Website crawl: structure, content and technical signals — no credentials needed. Google Search Console: real queries, clicks, impressions, CTR and positions for your verified property. Google Analytics 4: sessions, users, channels and key events. PageSpeed Insights: Google’s official lab scores and real-user Core Web Vitals. SERP provider (DataForSEO): ranked keywords, SERP competitors and keyword demand for your target market. Repository (GitHub): commit, release and CI cadence. App stores: public listing data — rating, rating count, update recency. Every credential is encrypted at rest (AES-256) and never shown back, exported or included in reports; the “?” on each integration card walks through its setup.
What we deliberately do not do
We do not estimate traffic, revenue, conversion rates, keyword difficulty of our own invention, competitor traffic or app downloads. We do not predict how much a fix will gain. We do not fill gaps when a source is disconnected. We do not recommend content for demand that only shares a generic word with your product, and we never surface unsafe demand as an opportunity regardless of volume. When this platform shows a number, a source produced it — that is the entire product.