Experiments
To know whether the cards actually increase sales, a store shows them to half its visitors and compares. This page works out how big that test must be, and lets you rehearse reading the results with simulated shoppers. Nothing here is a measured result; this demo has no real traffic.
1 · How long would a real test take?
Pick the smallest improvement worth detecting. Standard test settings: 95% confidence, 80% power, half of visitors see the cards (two-sided two-proportion z-test, intent-to-treat).
- Baseline conversion
- 14.8%
- Visitors per group
- 9,406
- Duration
- 65 days
Baseline is the conversion in this store's data. Smaller improvements need far more traffic: halve the improvement and you need about 4× the visitors.
2 · Rehearse the results with simulated shoppers
Runs your applied card settings click by click against simulated shoppers. How much each card helps is your assumption, set below: the chance a matching card settles a worry that wouldn't have settled on its own.