UX research can act as observations to form an initial hypothesis, while A/B testing can acquire the necessary data to perform experiments in way that assimilates the scientific method.

It might be a scale issue for sure, but qualitative data alone is not enough, especially when dealing with user biases and small, non representative populations.