Why it works

A finding is only worth what its methodology can defend. Ours is built so every claim in a report can be traced, re-run, and checked — by you.

We test the real paths, not the marketing copy

We don't review your homepage and guess. We enumerate the actual paths a customer or an agent uses to reach you — contact forms, booking widgets, hours/pricing pages, phone trees, chat widgets — and exercise each one with a real, realistic request (e.g. "do you have a 4pm slot Thursday," "what's your emergency callout rate").

  • Discovery surfaces — can a prospect (or agent) tell what you do and where, at a glance?
  • Information surfaces — hours, pricing, availability: is it published as text/structured data, or trapped in an image or PDF?
  • Action surfaces — contact forms, booking flows, phone/chat: can a real question or request actually be completed?

We test live, not from a transcript

Every finding comes from actually submitting the form, actually calling the number, or actually asking an AI assistant the question your customer would ask — not from reading your site and guessing what would happen. If we didn't do it, it's not a finding.

What counts as a failure

A failure is binary and observable: the form didn't accept the question, the page returned a 404, the assistant couldn't find the hours, the booking flow dead-ended. We don't score "vibes" or aesthetic opinions — every finding maps to something that did or didn't happen during a real attempt.

Versioned and reproducible

Test scripts and the surfaces we check are versioned and never silently changed. Every finding records exactly which test was run, against which version of your site, on what date — and we retain the raw evidence (screenshots, transcripts, submitted forms) behind every finding, so any claim in a report can be re-run and audited.

Measuring improvement

When we say a fix closed a gap, we prove it with a before/after re-run of the exact same test against the exact same surface. We label this what it is — an improvement measurement under controlled conditions — and never dress it up as a guarantee of leads or revenue.

What we will never do

  • Report a finding we didn't actually reproduce live.
  • Change a test silently and let results drift without a version bump.
  • Report a finding we can't back with retained evidence.
  • Guarantee leads or revenue from a check finding.