Skip to main content
Sometimes you just can’t run an A/B test at a per-user level. Some common causes are:
  • You don’t have per-user control over who sees what, e.g. a third-party marketing platform like Meta or Google Ads.
  • You can’t attribute metrics to individuals, e.g. total store foot traffic can’t be tied to which users received your email campaign.
Geotests help solve these problems. One of the most common use cases is marketing; you can’t A/B test on 3rd party ad platforms, but geo-testing lets you evaluate the incrementality of your program. Geotests are specialized experiments that allow you to target users based on geographic location. These tests enable you to test features, content, or experiences across any type of geographic area you specify. By splitting your campaign up amongst DMAs or postal codes, for example, you can measure its incrementality even when an ads platform doesn’t support it. Geotests are particularly valuable for businesses with global audiences who need to optimize their product for regional differences in user needs, preferences, or regulatory requirements.

​
Why use Geotests

Statsig’s Geotesting is designed to combine the flexibility and statistical rigor of statsig with your own data sources in your warehouse. Geotesting:
  1. Uses the data already in your warehouse, avoiding expensive and time-consuming exports.
  2. Includes an automated Experiment Designer to help you select where and when to run your campaign.
  3. Help export your campaign definition to your campaign platform, making setup faster and less error-prone.
  4. Automates analysis of the campaign, providing statistical rigor to measure incrementality.

​
GeoLift

Statsig’s Geotesting is built on top of Geolift, a best-in-class industry tool for running geospatial experimentation for marketing. GeoLift is an open-source package from Meta that’s used by many to infer causal relationships in timeseries data. It builds on advancements made in prior packages like Google’s Causal Impact package. Statsig is using GeoLift in our own implementation of Geotesting, and allowing you to easily connect your own internal metrics and feed them into the analysis package.

​
References

⌘I