PRE-REGISTRATION / DESIGN

Know what the experiment needs.

Set a meaningful effect, choose the error budget, and plan the sample before you start.

Design parameters

BINARY OUTCOME

Two-sided, independent proportions. MDE is an absolute percentage-point change.

Baseline conversion

Minimum detectable effect

Significance α

Target power

Treatment allocation

Eligible units per day

units

Normal approximation. Independent units, fixed horizon, no continuity correction. CUPED savings are not assumed.

REQUIRED SAMPLE

6,510control units

Treatment 6,510Estimated duration 14 days

Planning assumptions: 5.0% alpha, 80.0% power, 2 percentage-point MDE.

Power grows with the sample

The target is fixed before the outcome is observed.

1 pp MDE2 pp MDE3 pp MDE

◈ Calculated planning curves · normal approximation

CROSS-PROJECT CHECK / PRICEPOINT

Reproduction evidence

The repository tests the earlier pricepoint calculation against its documented assumptions. See the published calibration record for the reproduction and any missing variance assumptions.

test_reproduces_pricepoint_power