# What is a geo lift test?

> A geo lift test turns ads on in some cities and off in similar ones, then compares sales. It measures what ads truly cause, with no special tools.
- **Author**: Marxx
- **Published**: 2026-10-26
- **Category**: Measurement
- **URL**: https://marxx.ai/library-glossary/geo-lift-test

---

A geo lift test measures what your ads cause by running them in some places and pausing them in similar places. You then compare sales between the two groups of places. It works without tracking individual people, so it does not depend on cookies.

## How to run a simple one

1. **Pick matched cities.** Choose pairs with similar past sales, such as Pune and Ahmedabad.
2. **Split them.** Keep ads on in one city of each pair. Turn them off in the other.
3. **Run 3 to 4 weeks.** Short tests are easily fooled by one good weekend. Meta's [GeoLift guide](https://github.com/facebookincubator/GeoLift/blob/main/website/docs/Best%20Practices/BestPractices.md) says at least 15 days with daily data, and at least one full buying cycle (checked 2026-10-20).
4. **Compare orders** by city, against the same weeks before the test.

## A worked example

Before the test, two groups of cities sold about the same: 1,000 orders a week each.

During four weeks:

- Ads on: 4,600 orders
- Ads off: 4,000 orders

Lift = 600 extra orders over four weeks, or 15%. If ads in those cities cost INR 3,00,000, each extra order cost INR 500. Compare that with the cost per purchase your ad platform reported for the same cities.

## Where it goes wrong

- **Unmatched cities.** A city with a local festival in the test window will skew results.
- **Spill-over.** People in an "off" city still see ads while travelling, or from national campaigns.
- **Too small.** With only a few dozen orders per city, normal noise hides the effect.

## FAQ

### Is a geo test better than a holdout test?

They answer the same question in different ways. A [holdout test](/glossary/holdout-test) splits people. A geo test splits places. Geo tests can also capture offline and marketplace sales.

### Can I test one channel at a time?

Yes. Pause only Meta, or only YouTube, in the test cities, and keep everything else the same.

---
- [More Measurement articles](https://marxx.ai/library-glossary/category/measurement)
- [All articles](https://marxx.ai/library-glossary)