A geo experiment measures the effect of marketing by using regions, such as states or metro areas, as test and control groups. Advertising is paused, increased or launched in the test regions while the control regions stay unchanged. Comparing total sales in both groups against their historical relationship shows how much the change actually added or removed.
Because it uses total sales by region, a geo experiment does not depend on cookies, clicks or platform attribution. Open-source tools such as Meta’s GeoLift can help select regions and estimate how large an effect the test can detect.
Why it matters for eCommerce
Ad platforms report their own attributed conversions, which often overlap and overstate impact. A geo experiment gives an independent estimate of what a channel adds to total sales, which helps set realistic targets and budgets.
How it works in practice
- Define the question and main metric, such as total orders or new customers.
- Choose test and control regions with similar historical sales patterns.
- Run for at least 3 to 4 weeks, avoiding unusual periods.
- Keep other marketing the same in both groups.
- Measure results with store orders by region, not platform data.
Common mistakes
- Running tests that are too short or too small to detect an effect.
- Choosing regions that behave differently even before the test.
- Letting ads leak into test regions through other campaigns.