I'm thinking about a problem where I have data on different A/B tests and which variant won. The features would be a bunch of one-hot encoded functionalities of each page (i.e., blue button, square button, video on the right side, etc).
I want to see which features are most important to predicting whether a variant has won. Some tests can have multiple features (multivariate testing).
Does anyone have any ideas, giving this framework, to find out which features or combination of features are most important to figuring out what functionalities on a page are most important to predicting a winning variant?
In a test where you are testing multiple things (such as button size, button color, and text). How do testers usually tell which of the changes are most important?