As an intro - bumped into this opening paragraph from Liberty’s substack. I do like it and tend to agree with the Amazon approach: do lots of small experiments, then put more time and energy behind those that work.