Measure what an ad platform really adds

An incrementality test measures what an ad platform really brings in, by comparing sales with and without its ads. ProfitPilotPro offers two, in Marketing mix (MMM) → Tests & calibration: a geo test prepared and calculated by ProfitPilotPro, and the import of a Meta Conversion Lift result.

The geo test

During the test, Meta ads are switched off in a few regions (the control regions) and keep running elsewhere (the test regions). ProfitPilotPro reads your orders' revenue by shipping location (French region from the postal code, country outside France) and builds a synthetic region: a weighted average of the control regions that tracked the test regions as closely as possible over the previous 90 days. During the test, the gap between the test regions' actual revenue and this synthetic region is the revenue the ads brought in; divided by the test's Meta spend, it gives the incremental ROAS, with a 90% interval.

Prepare and run the test

1. 1ProfitPilotPro suggests control regions that make up 20–40% of your revenue and whose trend follows the rest of your sales. Tap a region to move it to the other group.
2. 2Choose the first day, the duration (14 to 56 days) and the daily Meta budget. The feasibility report gives the minimum detectable effect: the smallest gain the test would pick up 8 times out of 10. If it's above the expected ROAS, the test is flagged “likely inconclusive”; if your sales are too rare, “not enough volume”.
3. 3In Meta Ads Manager, exclude the control regions from the targeting of every ad set (Audience, then Locations) before the first day, and keep them excluded until the last one. Keep a steady budget.
4. 4The day after the last day, come back to the test and calculate the result.

Import a Meta Conversion Lift test

Meta Conversion Lift compares people who see your ads with a randomly drawn control group who don't. Copy the study's incremental ROAS, its bounds and the interval level (80, 90 or 95%). Access depends on Meta: its developer documentation states that Conversion Lift measurement is limited and goes through a Meta representative.

Calibrate the marketing mix on your tests

In Marketing mix (MMM), the Use tests as priors option (off by default) replaces a platform's starting point with the incremental ROAS of its most recent completed test; the width of the interval gives the margin of error. Each platform's card says which starting point is used. Meridian (Google) and Robyn (Meta) also calibrate on incrementality tests.

Limitations

An order without a country, or shipped in France without a readable postal code, isn't placed in any region. The department comes from the first two digits of the postal code: a town served by a post office in a neighboring department can end up in the neighboring region. All your active stores are read together, as in the marketing mix. The interval doesn't account for the uncertainty in the synthetic region's weights. The geo test only covers Meta.