It really depends what you mean by "low". It's certainly possible at the hundreds-per-day type volume. The key is to just look for bigger effects. If 200 people show up, 10/100 convert on A, and 25/100 convert on B then B wins (p < .01)

As your traffic scales, you can look for smaller and smaller effects. But the whole idea of testing variants can be applied pretty damn early.