Geo holdout

A geo holdout is an experiment that withholds or changes advertising in randomly chosen geographic areas and compares their sales or conversions with areas that ran as usual, to measure the campaign's incremental effect.

Google researchers described the method in 2011. Non-overlapping regions are randomly assigned to a control or a treatment group, and geo-targeted advertising delivers each region's condition. The authors found geo experiments conceptually simple, with a systematic design process and results that are easy to interpret (Vaver and Koehler, Google Research, 2011).

Geo tests need no cookies or user IDs, so they work for TV, CTV, out-of-home and channels where user tracking is weak. They need care in design: regions should be matched on past sales, the test must run long enough to detect the effect, and people who shop across region borders blur the result. Results also feed other models. Google's Meridian GeoX turns geo experiment results into priors for its marketing mix model (Google Meridian, accessed 9 November 2026).

Example: A grocery chain picks 40 comparable metro areas and randomly assigns 20 to keep a CTV campaign and 20 to go dark for six weeks. Sales in the exposed areas grow 3.1% more than in the holdout. The chain divides the extra gross margin by the media cost to get the incremental return.

Related terms: incrementality, iROAS, MMM, MTA, closed-loop measurement, CTV.

Authoritative source: J. Vaver and J. Koehler, "Measuring Ad Effectiveness Using Geo Experiments", Google Research (2011).