Choosing experiments one at a time based on results so far, rather than planning all experiments upfront.