That's true of all A/B testing for real world things isn't it?

You could have measured increased traffic because something was trending on twitter/HN instead of the UI changes you just deployed. Might be completely different tomorrow.