The hero lacks any trust signals like review count or rating, so first-time visitors may doubt comfort claims.
Current state
h1: Wildly Comfortable. Super Natural.; cta: SHOP MEN / SHOP WOMEN; notes: No social proof in hero section.
Change
h1: Wildly Comfortable. Super Natural.; cta: SHOP MEN / SHOP WOMEN; notes: Add a line under hero: 'Rated 4.7/5 by 50,000+ customers' or a small review widget.
Acceptance criteria
GIVEN a first-time visitor on the affected page
WHEN the change is live
THEN Add a line under hero: 'Rated 4.7/5 by 50,000+ customers' or a small review widget,
with no regression at 375px and 1280px widths.
Tracking
event: cro_no_social_proof on interaction with the
changed element; verify it fires in BOTH variants before opening traffic.
Success metric
conversion rate + bounce rate on the affected page
QA checklist
Chrome / Safari / Firefox · 375px + 1280px ·
keyboard reachable · screen-reader name present · no CLS introduced
Effort
Medium —
2-5 days
Test plan
299,882 visitors per variant
(95% significance, 80% power);
not testable at current traffic.
Rollback
Feature-flag the change; revert the flag if the primary metric
drops for 3 consecutive days.
annual visits × baseline conversion × relative lift × order value
= 2,160,000 annual visits x 2.10% conversion x 1.0-5.0% relative lift x $78 order value
The 1.0–5.0% relative-lift band is an assumption bounded by severity (High), not a prediction.
Why the range is conservative: Across 127,000 experiments, 12% produced a statistically significant improvement on the primary metric; a healthy programme win rate is 10-30%. Winning tests also overstate their true effect, so treat this as the prize if the change works — never expected revenue.
How effort is estimated
Bracket
Days
Typical scope
Low
0.5–2
copy, colour, labels, image swaps
Medium
2–5
layout shifts, form changes, new modules
High
5+
personalisation, funnel work, integrations
Assigned by matching the recommended change against these scopes; the estimate is a bracket, not a quote.
Sample size math
Two-proportion z-test, the standard pre-test sample-size calculation:
n = [z₀.₀₂₅·√(2·p̄·(1−p̄)) + z₀.₂₀·√(p₁(1−p₁)+p₂(1−p₂))]² / (p₁−p₂)²
Baseline conversion
2.1%
Significance / power
95% / 80%
Sample per variant, 5.0% lift
299,882
Sample per variant, 1.0% lift
7,353,946
Halving the minimum detectable effect roughly quadruples the sample required, which is why the conservative figure is so much larger.
What PXL is
PXL is CXL's prioritisation framework: binary questions instead of subjective 1–10 guesses, weighted toward evidence.
Findings discovered by inspection score zero on all four evidence questions — that is the honest signal, not a defect. Findings verified by direct measurement earn the extension point and, where real-user field data is involved, the analytics point.