On Design and Analysis of Funnel Testing Experiments in Webpage Optimization

In webpage optimization, funnel testing is a popular technique for studying the effects of webpage features. This study focuses on how to properly design and analyze the funnel testing experiments. To address the problem, we propose two new design construction methods for funnel testing experiments. The proposed methods fully utilize the underlying funnel structures such that the projected designs for different funnels have desirable properties. For the analysis of funnel testing experiments, we consider the penalized regression with the heredity principle to obtain parsimonious and interpretable models for each funnel. Thus, the optimization of finding the optimal settings of webpages can be appropriately achieved. The performance of the proposed design and analysis is illustrated through a few numerical examples.