EXP-014Growth team · analyzed Oct 7

Checkout test #14: one page beats three steps

One-page checkout (B) against the current three-step flow (A), 14 days, every checkout visitor.

DecisionShip variant B to everyone

Conversion up 10.0% with no hit to order value or refunds.

+10.0%95% CI +2.9% to +17.1% · p = 0.006
DatesSep 22 – Oct 6
Split50 / 50, by user
Visitors49,718
Primary metricCheckout conversion
AControl

Three-step checkout

6.00%1,489 orders / 24,812 visitors
+10.0%relative lift for B
WinnerBVariant

One-page checkout

6.60%1,644 orders / 24,906 visitors

How sure are we?

Relative lift with 95% confidence intervals. Anything that stays right of zero is a real improvement.

All visitors
+10.0%+2.9 to +17.1
Mobile
+18.4%+7.8 to +29.0
Desktop
+2.4%-7.2 to +12.0
-10%0+10%+20%+30%

Guardrails and secondary metrics

Nothing we care about got worse.

MetricABChangeResult
Average order value$58.20$57.90−0.5%No real change
Refund rate (14 days)2.1%2.0%−0.1 ptsNo real change
Payment errors0.4%0.4%0.0 ptsNo real change
Checkout load time, p751.8 s1.5 s−0.3 sBetter

By segment

The gain comes mostly from phones and first-time visitors, where three steps meant three slow page loads.

Mobile+18.4%
A
4.90%
B
5.80%

14,391 vs 14,445 visitors

Desktop+2.4% · not significant
A
7.52%
B
7.70%

10,421 vs 10,461 visitors

New visitors+15.1%
A
4.60%
B
5.30%

15,383 vs 15,442 visitors

Returning+5.4% · not significant
A
8.28%
B
8.73%

9,429 vs 9,464 visitors

Next steps

Oct 8Roll out to 100%

Behind the existing flag, so we can switch back in one click.

Two weeksWatch refunds

Refunds lag orders; re-check the guardrail on Oct 22.

Next testExpress pay on top

Test #15: wallet buttons above the address form on mobile.

Example from the MintPage showcase · fictional company, sample data · any agent can publish a page like this with one publish_page call