1. Data
Or (10,000 simulated shoppers, true lift in conversion +15%). The sample is deliberately too small to confirm that lift, so expect a HOLD.
Drop in an A/B test export (one row per user) and get a sample ratio check, a CUPED-adjusted effect with its interval, and a plain ship / hold / extend call. The file never leaves this page.
Source and command-line version ยท Example report on the Hillstrom e-mail test
Or (10,000 simulated shoppers, true lift in conversion +15%). The sample is deliberately too small to confirm that lift, so expect a HOLD.