AB testing aids enterprise operators with their determination making, and is taken into account the gold commonplace technique for studying from information to enhance digital consumer experiences. Nevertheless, there may be often a spot between the necessities of practitioners, and the constraints imposed by the statistical speculation testing methodologies generally used for evaluation of AB checks. These embrace the shortage of statistical energy in multivariate designs with many elements, correlations between these elements, the necessity of sequential testing for early stopping, and the lack to pool data from previous checks. Right here, we suggest an answer that applies hierarchical Bayesian estimation to handle the above limitations. Compared to present sequential AB testing methodology, we improve statistical energy by exploiting correlations between elements, enabling sequential testing and progressive early stopping, with out incurring extreme false optimistic threat. We additionally display how this technique may be prolonged to allow the extraction of composite international learnings from previous AB checks, to speed up future checks. We underpin our work with a stable theoretical framework that articulates the worth of hierarchical estimation. We display its utility utilizing each numerical simulations and a big set of real-world AB checks. Collectively, these outcomes spotlight the sensible worth of our method for statistical inference within the know-how trade.