Analytics
Pause or scale: decide with ROAS, delivery proof, and model disagreement
ROAS alone is not a pause button
Media buyers live inside campaign ROAS. Under 0.5× on last-touch looks like an easy pause. Often it is. Sometimes it is a delivery failure, a consent skip, or a model that only credits the last click while Meta did the early work.
Useful ad commentary starts with three numbers on the same window: attributed revenue, ad spend, and verified delivery for the destinations that should have received the conversion.
A practical pause / scale checklist
Run this before you change budget. It takes minutes when capture, spend, and delivery already live in one report.
- Confirm spend and revenue share the same date range and lookback window
- Open delivery proof for purchase (or your conversion) — failed or timed-out Meta CAPI sends inflate “bad ROAS” that never reached the auction
- Compare last-touch vs linear (or U-shaped) on the same journeys — if Meta collapses on last-touch but holds on linear, you may be cutting mid-funnel demand
- Scale only when ROAS clears your floor and delivery acceptance is healthy on the primary destinations
- Pause when ROAS is weak and delivery is clean — that is true creative or audience waste, not a logging gap
When models disagree, do not average them
Last-touch and linear often disagree on Meta because Meta sits mid-journey while branded search or email closes. Averaging the two into a “compromise ROAS” hides the decision you actually need.
If linear lifts Meta and branded search absorbs last-touch credit, protect Meta with a mid-funnel budget rule and judge search on assist-aware floors — not a single blended number.
What Intently makes inspectable
Intently keeps spend, multi-touch credit, CRM closed-won, and destination delivery on the same first-party journeys. Attribution Ask can surface campaigns under a ROAS floor with citations to model, window, and spend source — so pause and scale calls stay auditable.
If your stack cannot show delivery next to ROAS, fix that before you trust another optimization sprint.