Geo incrementality

What it is

Regions such as cities, postcode areas, states or media markets are assigned to test and control groups. One group receives a media change: switched off, switched on, or spend raised or lowered. The difference in outcomes between the groups, adjusted for pre-period differences, is the estimate of incrementality.

Why it matters

Platform reported conversions include demand that would have arrived anyway, so they can overstate the value of some channels. Geo tests give a privacy safe estimate that works consistently across search, social, video, audio, out of home and retail media. They are also the standard evidence used to calibrate mix models and to settle internal disputes about channel value.

How it works

Teams choose balanced region groups from historical outcome data, calculate the minimum detectable effect for the proposed budget and duration, then run the change while keeping control areas uncontaminated. Analysis uses difference in differences, matched market or synthetic control methods, and results are reported as lift with a confidence interval alongside an incremental cost per acquisition.

When it applies

It applies when a channel can be targeted by geography, regional outcome data exists, and the business can tolerate a holdout long enough to detect the effect, usually several weeks.

Examples

• Turning off brand search in a third of regions for four weeks to see how much of that traffic arrives through organic instead.
• Running an out of home campaign in three cities only and comparing store visits with matched cities.
• Increasing retail media spend by thirty per cent in selected regions to estimate marginal return before a national rollout.

How it is measured

• Incremental lift percentage with its confidence interval
• Incremental cost per acquisition or incremental return on ad spend
• Pre-period fit between test and control regions
• Minimum detectable effect achieved versus planned