Skip to content
GetProfit
Log in

← All terms

Geo lift test

A geo lift test is an experiment that switches ads off or raises ad spend in some regions but not in others, so the gap in sales shows what the ads really add.

How it works

Regions are split into test and control. In test regions, ads are switched off or raised (“go-dark” and “heavy-up” designs in Google’s GeoX); in control regions, they run as usual. Pre-test history gives the usual test-to-control sales ratio. Applied to control sales during the test, it gives expected sales; the gap to actual sales is the incrementality of the ads. The test reads the store’s orders, including those without an ad click.

A store needs regions that can be targeted separately and sell alike, with other marketing held equal. Meta’s GeoLift guidance recommends daily data, 20 or more regions, history 4–5 times the test length and at least 15 days of testing. Location targeting set to Presence skips people elsewhere who only show interest in a region. What a small store can run: incrementality testing for online stores.

Formula

Gap = actual test-region revenue − control-region revenue × pre-test ratio

In a go-dark test, a negative gap is what the ads were adding.

Example

Example store, not client data.

The tableware shop’s test regions usually sell half as much as control. For four weeks it switches the ads off there, saving 10,000. Control sells 240,000, so test regions should sell 120,000; they sell 105,000. The ads were adding 15,000: an incremental ROAS of 15,000 ÷ 10,000 = 1.5. Google Ads usually credits those regions with 45,000 for that 10,000: a ROAS of 4.5.

Not to be confused with

  • Campaign experiment — splits one campaign’s traffic and budget inside Google Ads. A geo test splits the map and reads the store’s sales.

Right and wrong readings

  • Wrong: “We switched the ads off in a region in July and its sales fell, so ads work.” Right: without control regions over the same weeks, a seasonal dip reads as an ad effect. July was among the three weakest months of ad revenue for 38 of 96 stores (GetProfit data, July 2025 – June 2026).

Sources