Multi-Arm Bandits with Design of Experiments to Continuously Optimize Websites

A/B Testing and Multi-arm Bandits are very often talked about as tools for analyze and direct traffic to website variants to brings in more sales. Multi-arm Bandits is touted to be faster, with lower lost opportunity compared to A/B Testing. What is interesting is how many variants should one test to get the best variant for highest sales possible? One approach is to use Design of Experiments (DoE) to understand interactions between variables. DoE allows minimal runs, or variants to understand how different variables like model, pose, clothes, background, etc... affect sales. From the interactions, you can not only choose the best variant to maximize sales, but also one that minimizes sales variability.