FIG 0.1
Findings, not opinions.
Each page is captured as a screenshot, its markup, its copy and its metrics. Only findings that cite one of them are scored.
How it works
A fixed pipeline that runs the same way every time. Every finding has to survive validation before it reaches your report.
The pipeline
Find the pages that matter for your goal: entry, pricing, signup, checkout.
sitemap + links · 5 pages
Load each page in a real browser, on desktop and on a mid-range phone.
1440 px · 390 px
Classify every page and its role in the decision the visitor is making.
pricing · role: compare
Run 32 published rules across copy, UX, UI, funnel, trust and performance.
41 candidates
Check every quote and element against the page. No proof, no finding.
24 rejected or merged
Compute the score, rank by impact and pin each leak on the screenshot.
17 findings · score 54
Confidence
Structural facts counted on the page, like missing labels or hidden CTAs.
Promises on one page checked against the page they lead to.
Copy read for outcome, audience and objections, quoted word for word.
Hierarchy, contrast and fold, shown as regions on the screenshot.
LCP, CLS, TTFB and failing scripts, measured on the pages of your conversion path.
Quotes must be found on the page and elements must resolve, or the finding is dropped.
Methodology
FIG 0.1
Each page is captured as a screenshot, its markup, its copy and its metrics. Only findings that cite one of them are scored.
FIG 0.2
Severity, confidence and reach set each leak's weight. The weights never change between runs, so the score only moves when your site does.
FIG 0.3
If a page times out or a category can't be checked, it is left out of the score and listed, instead of being counted as perfect.
Guardrails
Example data
The conversion audit where every finding comes with
Each leak points to an element on a real screenshot.
Copy findings cite the exact text on your page.
Speed and layout checks show the number and the threshold.
A finding without evidence never reaches your report.
The score is computed, so reruns never drift.
FAQ
The score is calculated from the findings with fixed weights, so an unchanged site returns the same number.
About a minute for the first results. Larger sites keep crawling in the background while you read.
It costs that page, not the run. The page is listed as not analysed, with the reason.
A category could not be evaluated in this run. It is excluded from the score instead of counted as perfect.
No. It reads public pages the way a visitor does. Nothing to install, no login needed.
Paste your URL. Leakly opens it like a first-time visitor and shows you what to fix first.
Free · No signup · Public URL only
Cookies
Strictly necessary storage keeps the site secure and remembers this choice. Analytics and ad measurement run unless you refuse them here. Cookie policy