Have it both ways : from A/B testing to A&B testing with exceptional model mining

摘要: In traditional A/B testing, we have two variants of the same product, a pool test subjects, and measure success. randomized experiment, each subject is presented with one variants, success aggregated per variant. The variant product associated most retained, while other discarded. This, however, presumes that company producing products only has enough capacity to maintain variants. If more available, then advanced data science techniques can extract profit for from testing results. Exceptional Model Mining such technique, which specializes in identifying subgroups behave differently overall population. Using association model class EMM, find subpopulations prefer A where general population prefers B, vice versa. This technique applied on StudyPortals, global study choice platform ran an design aspects their website.