Startups that are still searching for product/market fit don't really do much A/B testing - the user populations are too small for anything statistically relevant. They rely more on intuition, or as some would call it, blind luck.

In some ways, what an entrepreneur does is very coarse-grained A/B testing anyways. If you launch 50 different startup ideas and they all get zero traffic, and then you launch the 51st and it has 100 users sign up over a weekend, then there's a good indication that you should go with the 51st idea.

It's also what venture capitalists do at the level of companies. Fund a hundred companies, and then watch them. Pour more money into the ones who appear to be succeeding.