Platform-reported ROAS ignores returns. A product with 4× reported ROAS and a 30% return rate is not a winner — but nothing in your ads manager will tell you that. Return-adjusted net ROAS is CollectionsIQ's headline metric, and it comes in two flavors you choose per rule:
Net — settled. Subtracts returns that have actually happened. Precise, but it lags: recent sales haven't had time to be returned yet, so very recent windows look better than they'll end up.
Net — predicted. Forecasts expected returns for recent sales from your products' historical return behavior, giving you an early, self-correcting estimate. As real returns settle, the prediction converges to the settled number.
Attribution scope. When an ad drives an order containing several products, you decide how returns count:
- Full basket — returns of anything in the ad-attributed order count against the advertised product. Harsher, but reflects the real economics of the order.
- Ad-subject only — only returns of the advertised product itself count. Cleaner per-product signal.
Using it in rules. On any ROAS condition, set Return treatment to Net — settled or Net — predicted (default is Gross, the platform-reported value). A typical high-confidence rule: "net ROAS (predicted) at least 2.5, minimum $100 spend, last 30 days".
Requirements. Net treatments need your Shopify order and refund data flowing, which the app ingests automatically once installed. Note that net ROAS currently evaluates on rolling windows and store-wide data — market filters and seasonal periods apply to gross metrics.