Casual Inference Data analysis and other apocrypha

we’re always trying to both optimize and actually do something

explore vs exploit

example: subject line rotation

epsilon greedy

thompson sampling