A geo experiment is a method of measuring incrementality in which a territory is split into a test group and a control group of regions. The campaign runs in one group and is switched off or scaled down in the other, and after the period ends the development of sales in both is compared. Its advantage over tracking click-throughs is that it also captures the effect that never shows up in clicks – the customer who saw the advert and came in directly a week later. What matters most is the composition of the groups: the regions must be comparable in size, seasonality and past sales, otherwise the result tells you nothing. The test must run long enough to cover a typical purchase journey, and no other change, such as a sale or the opening of a branch, may interfere with it. Evaluate the result taking statistical significance into account, not merely the difference between averages.
See also: Incrementality, Statistical significance of a test, Media mix modelling.
