Measured, not promised
Here's how a number earns its place on this page, and why there are no borrowed case studies on it.
A result is control vs variant, on your own orders.
Every test keeps a control arm that receives no changes. Visitors are assigned by a sticky hash, so the same person always sees the same arm. Conversions are your orders (the Shopify pixel and order webhooks) or your conversion call, attributed to the arm that was live when they happened. The bandit moves traffic toward the winner while the test runs, but the control never drops below 10%.
Promoted only above 95% probability of beating control, with at least 200 sessions per arm.Pilot results land here as they clear that gate.
Each one will state the window, the sessions per arm, the control rate and the variant rate, the way the dashboard states them. Until a result clears the gate it isn't a result, so this page stays honest: no borrowed case studies, no modelled lift, no number without a control behind it.
Nothing on this page is an estimate.Why a held-out control, and not a before-and-after?
Because traffic changes. A launch, a sale, a new ad, a weekday: any of them moves a before-and-after number. The control arm sees the same traffic at the same time as the variant, so the difference between them is the variant. That's the only lift we'll put our name on.
Your conversion rate grows against a held-out control. Or you don't pay.
Your store could be the first result on this page.