General Sliced Factorial Designs for Online Experiments

Digital marketing is an integral part of digital transformation in industry.It is critical to use design of experiments to conduct online experiments for various forms of digital marketing including web design, email marketing, social network and others.Online experiments often involve multiple platforms, including desktop computers from different manufacturers, different types of mobiles and smart watches of different brands.A sliced factorial design is a suitable choice for designing online experiments with multiple platforms.We provide a general theory for sliced factorial designs and propose sliced generalized wordlength patterns to construct such designs for any number of platforms.The theory uses the characteristics of parallel flat design-based sliced factorial designs to construct optimal sliced factorial designs.For practical use, several constructed designs are provided in the Supplementary Material, enabling practitioners to efficiently design and analyze online experiments across multiple platforms.