Measure the funnel, not the clicks. A recommendation row earns its place when served products get added to carts and reach checkout, so the numbers to track are serves, clicks, add-to-carts, and checkouts, per row. Click counts alone flatter every widget and settle nothing.
The Funnel to Watch
- Serves establish the base: how often the row actually rendered with products in it.
- Clicks measure attention, useful mainly for comparing recipes against each other on the same placement.
- Add-to-carts are the first economically meaningful step: a served product entered a cart in the same visit.
- Checkouts close the loop: the visit went on to purchase after adding a served product.
Per-placement numbers matter because the same recipe can be great in the cart and mediocre on the home page. Judge each row on its own funnel.
Beware Generous Attribution
Recommendation widgets have a habit of grading their own homework kindly: any purchase within a long window after any impression claims credit, and the dashboard announces revenue the row never influenced. Prefer conservative rules, where a click counts only on a card that was actually served, an add counts only for a served product within the same visit, and a checkout counts only when the served product was added first. Dynasort measures every placement with exactly those rules, so a lift in the numbers is a lift in reality.
Dynasort measures recommendation rows conservatively, serves to checkouts. Install it from the Shopify App Store or see how Dynasort attributes results.