Six months on one store, and which of it was mine.

The clearest result of the three. The interval on conversion sits entirely above zero, so the effect is established rather than suggestive. Revenue per visitor moved with it but its interval just touches zero, and order value is flat, so the gain is more buyers rather than larger baskets. The subscription share of orders did not move either, so the extra buyers arrived on the same mix.

More people bought, and each spent slightly less. Conversion carries 98% probability to beat control, but the reported interval reaches down to zero, so it is probable rather than proven, and revenue per visitor and order value both span zero outright. The subscription share of orders did not move, which matters here: the lift came from more buyers on the same mix, not from pushing people off the subscription. The platform's summary also recorded abandoned cart and abandoned checkout rates both falling, which is the mechanism you would expect if the cue were doing its work rather than simply being noticed. Shipped on that basis, with order value kept under watch.

The topline never separated from control: conversion, revenue per visitor and order value all have intervals spanning zero at 66.1% probability to beat control, on 216 to 238 orders per arm over 16 days, which is simply not enough to call. What did separate was the unit economics. The platform's own summary put revenue per unit up 42% at close to full confidence on 27% fewer units per order, and the subscription share of orders fell 2.88%. So the higher price behaved exactly as a higher price should, earning more per unit from a slightly smaller and slightly less subscription-weighted basket, without measurably costing conversion. That is a reason to re-run at a larger sample, not a result to act on, so the price stayed where it was. Reported here because a testing programme that shows only winners is not a testing programme.