Method
The method is in the product, not in a help doc.
Any figure travels with its window, its sample, or its control. Here is all of it in one place.
01
Install and pick placements
Add the app from the Shopify App Store and choose where recommendations appear. Ten minutes, no theme edits.
02
A control group is held out
8% of sessions see your store without recommendations. They are the baseline every figure is measured against.
03
Revenue is attributed per widget
Orders are traced back to the widget and session that influenced them, inside a 7-day click-and-view window.
04
You check the number
Switch a widget off and the reported revenue moves. That is the test, and the product is built so you can run it.
Assumptions
Everything we assume, written down.
| Control group | 8% of sessions, randomised per visitor, sticky for 30 days |
|---|---|
| Attribution window | 7 days, click and view, last non-direct touch within the app |
| Revenue basis | Net of discounts and refunds, excluding shipping and tax |
| Reporting lag | Orders settle for 48 hours before they are counted as final |
| Minimum sample | Lift is withheld until 1,000 sessions per arm |