Split has increased some of our CVR; for example, in a recent pop-up model experiment, our conversion was around 15% and once we started the experiment, the on-variant won by increasing the CVR up to 35%, making it a really good tool for determining which particular feature would deliver the expected CVR.
