Experiment readout · 1 June–9 August 2026

F brings more visitors through to purchase

Across eight equally allocated page versions, the clearest positive signal is on the outcome that matters most: purchases per visitor. F appears to lead; C attracts far more clicks, but that extra interest does not translate into more purchases.

Recommendation: take F forward to a controlled rollout or confirmatory test against A. Keep A as the fallback until the result is validated against business guardrails. Do not select C based on its click rate alone.

What happened

F is the most promising candidate on the end-to-end result. Its purchase rate is approximately 2.2%, versus about 1.8% for the current page A—roughly a 0.3 percentage-point gain, or around one-sixth more purchases per visitor. With traffic split evenly, that is a meaningful directional difference, not just a consequence of F receiving more visitors.

Best purchase outcome F · ~2.2% Purchases per visitor
Current page A · ~1.8% Baseline for comparison
Attention ≠ sales C · ~49% click rate Highest click rate; purchase rate ~1.8%

Purchases per visitor: the decision metric

Approximate pooled rates, shown as a percentage of assigned visitors. A is the control.

Approximate purchase rate by page version F is highest at about 2.2 percent. A, B, C, D, E and G are near 1.8 percent. H is lowest at about 1.6 percent. 0%0.5% 1.0%1.5% 2.0% Rate ~1.8%A ~1.8%B ~1.8%C ~1.8%D ~1.8%E ~2.2%F ~1.8%G ~1.6%H

The chart is a rounded, directional summary. Small differences among A, B, D, E and G should not be read as a reliable ranking without formal uncertainty estimates.

The funnel explains why clicks alone would mislead

Rates below use the step-specific denominators in the experiment definition: clicks ÷ visitors, adds to cart ÷ clicks, and purchases ÷ adds to cart. They are rounded summaries.

Approximate funnel rates for A, C and F C has a much higher click-through rate, but lower add-to-cart per click and purchase per cart than F. F has the strongest overall purchase rate. Click / visitor Cart / click Purchase / cart Purchase / visitor A · Current C · Click leader F · Best outcome ~36% ~15% ~36% ~1.8% ~49% ~14% ~25% ~1.8% ~38% ~17% ~36% ~2.2% Bars are illustrative within each column; each funnel stage has its own scale.
C is a cautionary example: its unusually high click rate does not carry through the funnel. F gets more visitors to add an item after clicking, then retains a typical cart-to-purchase rate. Evaluate redesigns on completed purchases, not CTA clicks in isolation.

Rounded rate comparison

Version Click / visitor Cart / click Purchase / cart Purchase / visitor
A (current)~36%~15%~36%~1.8%
B~36%~15%~35%~1.8%
C~49%~14%~25%~1.8%
D~36%~15%~36%~1.8%
E~32%~15%~37%~1.8%
F Lead~38%~17%~36%~2.2%
G~35%~15%~35%~1.8%
H~34%~14%~34%~1.6%

Rates are rounded summaries of the aggregated experiment data; they are intended to show the broad pattern, not to imply precision at the displayed decimal level. Because the supplied data are daily aggregates, this readout does not provide visitor-level confidence intervals or a multiple-comparison-adjusted significance test.

What to do next

01 · Validate

Run F against A

Use a fresh, pre-planned test with purchase per visitor as the primary metric. Set the decision window and stopping rule in advance.

02 · Protect

Check business guardrails

Confirm revenue per visitor, margin, cancellations/returns, and page performance. More purchases are not automatically more profit.

03 · Learn

Diagnose the funnel

Investigate why C produces many clicks but fewer downstream conversions. Use that learning to improve intent and clarity, not just click volume.

How to read this result

Visitors were randomly assigned in equal proportions and retained their assigned version, which makes the comparison useful for estimating the effect of these page experiences. The primary decision outcome here is purchases ÷ visitors; the other rates help explain the path to purchase.

This is a product decision readout, not a formal significance report. The seven-candidate comparison creates multiple opportunities for a chance winner, and daily aggregates do not expose visitor-level behavior or uncertainty. Treat F as the strongest candidate to validate—not as proof that every redesign detail caused the lift.