Earlier quoted context omitted.
I blame this on A/B tests which ignore externalities. Sure, maybe the 3 metrics you were aiming for were improved, but your brand is fucked. I think brands like Sears/Kenmore/Craftsman were A/Bed to death. So many little 'optimizations' (read: cost cutting that looked-good-on-paper) that eventually destroyed what the brand was about in the first place.
I think this is related to my fundamental problem with A/B testing: when people A/B test individual variables they'll take a greedy optimization that gets stuck in local minima. The other problem is that you only understand the change in metrics you measure and things like "satisfaction" are often lost.
I think you wanted to say local maxima, but yes I agree.