Attach rate is computed as orders containing both A and B / orders containing A. The denominator is restricted to orders that already contain the anchor item, not all orders — that restriction is the whole point. Naming the pair in advance (“attach rate of grinders to whole-bean coffee”) is what separates attach rate from a basket-size average; without a named pair the metric collapses into a different question.
Operationally, attach rate is the primary metric behind bundle merchandising, PDP “frequently bought together” modules, cart upsells, and post-purchase one-click offers. A coffee brand tracks grinder-to-bean attach; a footwear brand tracks socks-to-shoes; a skincare brand tracks serum-to-cleanser. Each surface is a controlled experiment whose success is measured as attach-rate lift on a specified pair.
Attach work is one of the highest-ROI ecommerce levers because it raises AOV against an already-paid acquisition cost — the incremental line item rides on the anchor order’s CAC.
Distinct from Units Per Transaction (UPT), which counts total units regardless of composition: a 2.1 UPT describes basket size; a 38% serum-to-cleanser attach rate names a specific pairing that’s pulling. Pitfalls: small denominators are noisy, promotional bundles inflate measured attach without indicating organic affinity, and an absolute attach number is uninterpretable without a baseline — 12% is good or bad only relative to the pair’s history.