I’m pretty confident that a lot of this is basically because of AB tests. Even in a top app like Facebook it can take many months to hit statistical significan

I’m pretty confident that a lot of this is basically because of AB tests. Even in a top app like Facebook it can take many months to hit statistical significance on a new feature test. If the numbers on that test are down (I.e it’s not moving in the right direction) they don’t just give up on it, they iterate and try again. Add to this NUXs, not being diligent about deleting unused code and you have a receipe for a phat app.