“The hardest part is taking a loss in an experiment is not intuitive if you only always go by positive revenue metrics you will burn yourself to the ground.”

Taking a Revenue Loss in an Experiment Is Not Intuitive — But It Is Necessary

Jim's operating principle at Duolingo: accept that some tests will show negative revenue impact and ship them anyway if they serve the 100-year version of the company. Over-optimising purely for short-term revenue metrics compounds the premium trap — it feels good each quarter until the brand is hollowed out.