>>2076the "complete fluke" part is spot on. if you dont document the
pivot points , a client will just assume you got lucky with an algorithm change or a seasonal trend. i always try to include a section specifically for
failed experiments so they can see exactly what we ruled out before finding the winning path ✅