A/B testing, also known as split testing, in its most basic form comes down to two different implementations shown at random to different kinds of people. A small number, say 5%, are shown the new feature, A, which could be something like a new feature or a new view, and another 5% will see feature B. The remainder of the users will not see the new feature yet. The feature that will prove to be most popular (by conversion or otherwise, depending on the objectives) will be fully implemented and offered to the complete audience of your app.
In case you want to find out what works best for signing up users, you can set up a split test like this:
So 50% of your test audience sees variation A, showing a button that says Sign up, which will lead 26% of the visiting users to sign up. The other 74% might think "Hmm, this is not for me...