Geo test sample size calculator
Screen conversion volume before designing a geo holdout test. Enter weekly baseline conversions across the regions, the share of conversion volume held out, and expected relative lift. This simplified calculation is an initial planning aid. It cannot determine how many regions you need or replace a power analysis using historical regional outcomes.
The first listed duration passing this simplified volume screen is 12 weeks. This does not establish that your geo experiment has 80% power; validate the design against historical regional data.
Approximation assumes independent Poisson conversion counts, stable baseline rates and no regional correlation. It ignores geo heterogeneity, overdispersion, seasonality and model fit. It is not a region-count calculator or a final power analysis.
How to read the result
More conversion volume reduces the threshold in this simplified model. Passing the screen is a reason to investigate a design, not a promise that a test will detect the effect. Regional differences, correlated outcomes and spillover can materially change precision. Start with our free guide to geo experiments.
Formula and assumptions
Approximate relative lift threshold = (1.96 + 0.8416) × √(1 / treatment conversions + 1 / holdout conversions). The constants represent a two-sided 5% significance level and 80% power in a normal approximation. Both conversion counts are baseline weekly volume × allocated share × weeks.
For 250 conversions per week, a 20% holdout and eight weeks, the baseline counts are 1,600 and 400. The approximation gives 15.7%. At twelve weeks it gives 12.8%. An expected 15% lift therefore first passes the listed screen at twelve weeks. These are illustrative calculations, not simulated geo-test power.
What to do before launching a test
Choose targetable, sufficiently isolated regions; examine historical trends; check balance and spillover; and simulate the proposed analysis on pre-test data. Include the planned spend change, outcome lag, uncertainty method and stopping rule. Google’s Matched Markets methodology illustrates why geo design requires more than a total conversion count.
Once a completed test produces an incremental revenue estimate, use the iROAS and margin calculator to interpret its economics.