Blog
How do incrementality holdouts work on Meta?
How do incrementality holdouts work on Meta?
Incrementality tests compare exposed versus holdout to estimate lift. Meta and others offer conversion lift and geo designs. Read current product docs. Do not invent a 2026 average lift percentage; eonik does not run holdouts or touch spend; After you know whether media is incremental, you still decide what creative to make next; You approve that cut.
How do incrementality holdouts work on Meta?
Incrementality tests compare exposed versus holdout to estimate lift. Meta and others offer conversion lift and geo designs. Read current product docs. Do not invent a 2026 average lift percentage; eonik does not run holdouts or touch spend; After you know whether media is incremental, you still decide what creative to make next; You approve that cut.
Design before vanity lift
Pre-register the metric, window, and power. A noisy holdout is worse than no test. MMM is a different method, aggregated, not a substitute incrementality figure.
Meta’s geo-based MMM research is method, not a lift claim you should paste.
Do this in order
Run the test as a test. Do not automate spend off a noisy interval.
- 1. Pre-register metric, window, and minimum power with finance.
- 2. Choose conversion lift or geo design from current Meta docs.
- 3. Report the interval, not a vanity single number.
- 4. If media is incremental and the opening is still tired, make the next on-brand cut in eonik. You still click spend.
eonik never runs the holdout
No ad-account write. No incrementality product. Production only.
After the test, the lever you still control is the file.
Holdout vs MMM vs creative test
| Method | Grain | eonik? |
|---|---|---|
| Incrementality holdout | Exposed vs holdout lift | No. Never spends |
| MMM | Aggregated time series | No |
| Creative sandbox | Hook / body isolation | Makes the files you approve |
What usually breaks
- Inventing a 2026 average incrementality percentage.
- Shipping a noisy holdout without pre-registered power.
- Treating MMM as a substitute incrementality figure.
eonik is not an attribution, tagging, or budget suite. After you understand the readout, the lever you still control is what you make next. Start at how to generate AI ads. You approve every cut; nothing spends.
“Almost any question can be answered, cheaply, quickly, and finally, by a test campaign. And that’s the way to answer them, not by arguments around a table.”
That principle still holds. The next on-brand cut is yours to approve.
Sources
Questions
What incrementality percentage is normal?
There is no normal we will invent. Report your test’s interval. Meta and others offer conversion lift and geo designs. Read current product docs; MMM is a different aggregated method, not a substitute incrementality figure; eonik does not run holdouts or touch spend.
How do incrementality holdouts work on Meta?
You compare exposed versus holdout to estimate lift. Pre-register metric, window, and power. A noisy holdout is worse than no test. Use current Meta product docs for conversion lift or geo designs; Do not invent a 2026 average; After you know whether media is incremental, you still decide what creative to make next.
Can eonik run a Meta holdout?
No. Never spends, never writes to the ad account. After the test, consider the next on-brand cut, assemble it on Mac, approve it. Start at how to generate AI ads, then download; Incrementality design stays with you and Meta’s tools.
Free, no card · macOS 15+ · Apple silicon