Case study
We audited our own landing page. It hurt. We published it anyway.
fixlist grades pages against a strict 30-point conversion rubric — so the obvious test was to point it at page.screela.com itself. It gave us 64/100 and called our trust signals "nearly absent" (28/100). So we did what we tell our customers to do: implemented the fix list, re-ran the audit, and repeated — three rounds in two days. Final score: 75/100 — trust +30, readability +20. Every raw report is linked below, unedited.
Raw, unedited reports: round 1 — before (64) · round 2 (67) · round 3 (70) · final (75)
What the audit caught
16 findings. The verdict, verbatim:
"The hero nails the 5-second test… But trust is nearly absent: zero testimonials, logos, numbers, or a visible sample report despite it being in the nav, which is fatal for a paid product. Readability suffers from low-contrast gray body text throughout, and the persuasion arc skips proof entirely before asking for money."
What we shipped, round by round
The report said visitors had to imagine what they were buying. Now a real annotated screen from the sample report sits right on the landing page, numbered boxes and all.
“Persuasion arc skips proof before the ask” — so before pricing there is now a section with something rare: our own unflattering audit, published.
Body copy was below WCAG AA against white. Muted text went from #6b6560 to #4c4741 — same design, readable at a glance.
A privacy line now sits under the URL field, right where the money is asked for.
Consistent green CTAs on both pricing cards, a persistent “Audit my page” button in the header, alt text on the mascot, and a company identity line in the footer.
“The hero visual is decorative, not the offer” — the doodle gave way to a real excerpt card from the sample report, score, chips and a critical finding included; the mascot now just peeks over its corner.
The re-run called the question-form headline awkward and suggested the exact wording. We shipped its sentence verbatim: “Your landing page is leaking customers. Find out exactly where — in 90 seconds.”
#333 still read as gray to the judge, so body copy went to near-black with gray reserved for microcopy. Readability: 62 → 82.
“Self-deprecating copy undercuts confidence” — the proof section now opens with the measured improvement and links the evidence, instead of cataloguing our own sins.
What moved
| Score | Before | After | Δ |
|---|---|---|---|
| Overall | 64 | 75 | +11 |
| Trust | 28 | 58 | +30 |
| Readability | 62 | 82 | +20 |
| Call to action | 68 | 80 | +12 |
| First impression | 72 | 78 | +6 |
| Hierarchy | 70 | 74 | +4 |
| SEO | 88 | 88 | ±0 |
| Mobile | 75 | 75 | ±0 |
What it refused to give us
The re-run did not hand out a friendly score. It kept "no real social proof" as a critical finding and dryly noted that publishing your own audit "is clever but not third-party proof." Correct — testimonials can't be shipped in a commit; they have to be earned. That's the point of the product: the rubric doesn't care whose page it is, and a score you can't flatter is a score you can trust.