A/B testing

Offline metrics predict quality; an A/B test proves it on real traffic. Visitors are split into variants, each served a different ranking configuration, and metrics — click-through, conversion, zero-result rate — are compared with confidence intervals so decisions aren't made on noise.

AACSearch supports A/B/n experiments natively with sticky variant assignment, so a shopper keeps seeing the same variant across sessions.