Checkout: CartFlows against checkout v3
Running since 6 Oct 2026 · data to 11 Nov 2026 (37 days) · refreshed 12 Nov 04:02 UTC · radio-controlled.co.uk
Hypothesis. A checkout in the new visual system (v3) completes more orders than the CartFlows checkout it replaces.
Orders
CartFlows checkout
19.3%
64 sessions
Checkout v3
18.0%
61 sessions
-6.7%No significant difference yetp = 0.670
Totals by arm
| Arm | Started checkout | Users | Checkout sessions | Orders | Revenue | Revenue / session |
|---|---|---|---|---|---|---|
CartFlows checkout Control · checkout_v3_control | 305 | 305 | 332 | 64 (19.3%) | £11,483 | £35 |
Checkout v3 Variant · checkout_v3_v3 | 310 | 310 | 339 | 61 (18.0%) | £7,863 | £23 |
Orders are paid WooCommerce orders carrying the arm, as a share of sessions that started checkout in that arm (GA4).
Cumulative rate by arm
Share of all sessions to date that reached the step. Settles as the sample grows.
CartFlows checkoutCheckout v3
How to read this
- Significance is a two-proportion z-test on the session rate; it is only shown once both arms have at least 25 of the event. Before then the uplift is noise.
- A difference is worth acting on when it is significant on orders or revenue per session, not on product views alone.
- Visitors are assigned on their first load of /checkout/, including with an empty basket, so some who started checkout never had anything to buy.
- Started checkout comes from GA4 (visitors sent an impression on the checkout page). Orders and revenue come from WooCommerce: orders carrying the test's arm, counted when paid (processing, completed or on hold), net of refunds, in each order's own currency.
- This test does not set the GA4 user property, so product views and add to cart cannot be split by arm and are not shown.
- The assignment counters kept on the server (bash deploy/apply.sh report) are the authoritative denominator; GA4's count can run a little lower.
Data from GA4 property 307789325 via the daily snapshot Worker. Since 6 Oct 2026.