On November 12 Meta for Business published guidance on using incrementality, not just attribution, to decide where ad budgets go. It draws on closed-door sessions Meta held with finance, marketing and analytics leaders in North America and the UK. With Q4 spend peaking and 2027 budgets being set, it is a useful framework for anyone defending Meta spend to a client or a CFO.
The core argument
Meta draws a clear line between the two: attribution shows where a conversion was recorded, while incrementality estimates whether the ad actually caused it. The post argues that as AI speeds up spending and decision-making, causal measurement is the only way to tell whether more automated output produced value or just activity.
On Meta specifically, it points to Conversion Lift, which compares outcomes between people eligible to see ads and a randomised holdout group, and recommends using those results to calibrate attribution models and marketing mix models.
The four practices Meta highlights
The guide groups its advice into four habits it says separated incrementality-led teams from the rest.
- Agree with finance on which outcome defines value before budget season, not during the review
- Walk finance through the measurement method and its limits early, so it is challenged before it backs a budget request
- Test brand spend over a period that matches how brand builds value, and plan with a credible range rather than one number
- Start with one high-spend channel where last-touch and incremental results are likely to disagree, such as branded search
What media buyers can take from it
Platform-reported ROAS will always look best inside the platform that reports it. A holdout test gives you a number you can defend when a client asks whether Meta is driving sales or just claiming them. It also tells you how far to trust day-to-day Ads Manager figures between tests.
This is guidance, not a product change. Nothing in Ads Manager moved, and Conversion Lift availability and minimum requirements still depend on your account and spend.
If you only have budget for one test, Meta suggests a high-spend channel where last-touch and incremental results are likely to tell different stories. Its own example is branded search, which may simply capture demand that already exists. In many Meta accounts, retargeting raises the same question and is a sensible first candidate.
What it means for you
- Agree the success metric with the client or finance team before you plan 2027 budgets.
- Run a holdout or Conversion Lift test on your largest Meta campaign, then calibrate daily reporting against it.
- Report Meta results as a range with stated assumptions, not a single ROAS figure.
- Steady spend helps lift tests read cleanly; stable accounts with reliable billing make that easier.
