Measurement & tracking

Conversion math (per variant)

Daily aggregation of sessions (n) and conversions (k), conversion rate p = k/n; tests run to pre-set decision thresholds (Frequentist or Bayesian, as above).

Behavioral analysis on conclusion

On test end, compare pre-campaign (T₀-14→T₀) vs during-campaign (T₀→T₁) behavioral metrics from GA4 (engagement rate, scroll depth, bounce rate, etc.). This reveals not just who won, but how behavior shifted, guiding better next iterations. The report resets if the significance time frame is reset.

Status lifecycle

continue_testing → variant_winner / control_winner or no_conclusion (≥21 days) if thresholds never met. Resuming a paused test prompts a time-frame reset option for clean evaluation.