Most audits hand you a score and a list of warnings. We hand you evidence: the URL, the date, the exact code, and a fix you can paste in. This page is the process — published, because we hold ourselves to the same standard we hold your site.
01 The quality gate
Every report — for every client, every time — must clear a 100-point scorecard before delivery. This is not a marketing claim. It is a hard stop in our workflow.
Out of 100 points across evidence completeness, action clarity, technical accuracy, and plain-language quality. Below 90, the report goes back — no exceptions, no "close enough."
Certain errors kill a report instantly regardless of score: a missed noindex blocker, a security issue overlooked, invented data, or an unevidenced P0 claim. One hard-fail = do not deliver.
Before any report leaves, we re-crawl the site, spot-check 5 findings against live pages, and scan the full text for the word "guarantee" — because we never promise rankings, and we check that we didn't accidentally imply one.
Our own post-mortem (October 2026) documents how five independent review passes caught critical errors the original author structurally could not see — including a client-data leak and a pricing contradiction. The review loop is not optional process. It is the safety net.
02 How we score
Every finding is scored on five factors, then gated. A high score alone doesn't make something P0 — it must be a confirmed blocker.
The five factors
Impact (does it cost revenue or visibility?) · Risk (can it get worse or trigger a penalty?) · Evidence confidence (how sure are we?) · Scale (one page or the whole template?) · Readiness (can you fix it today?). Each scored 0–3.
The P0 gate
A finding is P0 only if it is a confirmed crawl, index, trust, or security blocker — something that demonstrably prevents a commercial page from being crawled, indexed, or served accurately. A schema price that contradicts the visible price is P0 (trust blocker). A redirect chain that resolves is P1 (weakness, not failure). The gate overrides the score.
The classes
P0 (13–15 + gate): fix this week — your store is actively losing money or trust. P1 (10–12): fix this month — significant drag on visibility. P2 (7–9): fix this quarter — meaningful improvement. P3 (0–6): note it, revisit later.
03 Evidence standards
You should never have to take our word for anything. Every statement in a SunmosLab report is tagged so you know exactly where it came from.
We fetched the page, saved the HTML, and recorded the date. The evidence file is in your report appendix. You can check it yourself — that's the point.
Facts only you can supply — registered address, shipping terms, supplier specs. Marked as client-provided so they're never confused with something we verified independently.
We couldn't test it — no access, no data, no reliable method. We say so plainly instead of estimating. A gap in the report is a finding, not a failure.
04 Radical transparency
In October 2026 we shipped a pricing page that contained a client-data leak, an internal placeholder, a pricing contradiction, and dead CTAs. Five independent review passes caught them before a prospect ever saw the page. We wrote a post-mortem so it cannot recur — and we're linking it here because a service that sells evidence should be able to produce its own.
Client data in a public template
A generic pricing page carried another client's crawl dates and URL counts, copied from a forked template. Caught in review. Rule created: any template derived from client work gets a mandatory sanitization pass — remove all client nouns, dates, URLs, numbers — before it ships.
"Insert your first result here"
An internal placeholder sat in the Proof section, visible to any buyer. Caught in review. Rule created: the final pre-ship check for any client-facing file is a placeholder sweep — one command, three seconds, zero tolerance.
The Lite tier made the flagship pointless
Three months of Lite ($450) included the $400 audit plus articles and re-tests — making Tier 1 strictly worse than the entry tier. Caught in review. Rule created: before any pricing artifact ships, we cross-check every tier for arbitrage. A rational buyer should never be able to route around the product we're selling.
Why we tell you this: because a service that promises "evidence, not promises" has to be able to show its own evidence — including the evidence of where it failed. The post-mortem is dated, the rules are standing, and the review loop that caught these errors is the same loop that will review your report before it reaches you.
05 What this means for you
You get a fix, not a lecture
Most audits end with "your schema is missing." Ours end with the corrected schema block, ready to paste, plus the URL it goes on and the date we checked it. Your developer — or you — can implement most fixes in an afternoon.
You know what we couldn't test
Where we lack access or data, the report says "Not measured" — not a guess, not an estimate, not a vague "needs improvement." You always know the boundary of what we verified.
You can check every claim
Every finding cites the URL, the date, and how to verify it yourself. If we say your product pages are missing schema, we show you the exact page and the exact missing code — and you can confirm it in two minutes.