Hacker News new | ask | show | jobs
by codeddesign 3699 days ago
A/B testing data is only as good as the samples provided. They may have done wonderful A/B testing, but that the alternative designs were just so bad that the current design version made it through ;)
1 comments

A/B testing is, fundamentally, hill-climbing. A/B testing will select the fastest horse, but never the strange wheeled box belching noxious fumes.