Lore
GeoLift test workflow
A five-step process, demonstrated with Recast's GeoLift, for running a Geographic experimentation test with Synthetic controls.
- Gather data. Daily revenue by geography, at least 90 days and ideally a year.
- Design the experiment with simulations. Specify the spend increase, metric, expected incremental ROI, duration and budget, then pick a top-ranked configuration.
- Run it. For example, +$15,000 in Austin, Boston and Denver for 4 weeks.
- Upload post-test data.
- Analyze. Compare actual revenue to synthetic-control (counterfactual) revenue and study the gap between the lines.
Design-stage insights
- Design tells you up front what the test can and cannot detect. For example, an 80% chance of detecting an effect if true incremental ROI is at least 3x. If the true effect is smaller, the result may be invisible or the uncertainty too wide to be useful.
- The ROI input must be incremental ROI, not platform-reported ROAS.
- Hold everything else constant in the test markets during the run.
Interpret results with Reading incrementality results through the confidence interval. Trade-offs are covered in Incrementality experiment design trade-offs.
Note: the source is a product-led pitch for Recast's GeoLift.