Learn by Directing AI
All materials

verification-targets.md

Verification Targets: Overall Conversion Rate Test

Overall Test Targets

These are the known-good values computed from the experiment dataset. Use them to verify your AI-generated statistical test results.

Metric Value
Test type z-test for two proportions (or chi-squared test for independence)
Page A visitors 2095
Page B visitors 2105
Page A bookings 356
Page B bookings 445
Page A booking rate 0.1699 (17.0%)
Page B booking rate 0.2114 (21.1%)
Difference (B - A) 0.0415 (4.1 percentage points)
z-statistic 3.4207
p-value (two-tailed) 0.000625
95% CI for difference [0.0177, 0.0652] (1.8pp to 6.5pp)
Effect size (Cohen's h) 0.1057

How to Use These

Compare your AI-generated results against these targets:

  • The p-value should match within rounding (same order of magnitude, same first two significant digits).
  • The CI bounds should match within 0.1 percentage points.
  • The effect size should match within 0.01.

If they don't match, check:

  1. Whether AI used the correct test type (z-test for proportions or chi-squared, not a t-test -- the outcome is binary)
  2. Whether AI applied the correct metric definition (booking_completed = true / total visitors per page version)
  3. Whether AI included all rows (no accidental filtering by tour type or visitor source)