What is it?
We run advertising in some locations and not in others, tracking the performance of KPIs to measure the difference it makes.
Geo-experiments are very different to simply running some activity and seeing what happens, they are robustly constructed, with prior analysis so that we know we have selected the right target geographies and the test has enough scale to generate measurable results. Once the campaign has ended, we use ‘synthetic counterfactual’ models to produce the most accurate performance read that we can, rather than simply comparing pre and post-campaign differences.
When is it used?
- Testing a new channel or creative
- Confirming critical results from marketing mix modelling before acting on them
- Improving the performance of marketing mix models by grounding them in known test results
- Producing simple, persuasive evidence of advertising’s effectiveness, without requiring trust in a complex statistical model
Why Sequence?
- Tests are robustly designed to generate results: which regions to test, which to hold out and how long to run
- Deep understanding of how media delivery and online KPI response data can spill outside of targeted regions, and designing tests to avoid this cross-region pollution
- Results come with honest uncertainty ranges
- Simple, impactful communication of uplifts to share with stakeholders
Next service
A/B testing →