I find it suspicious that they would use user engagement and retention and none of the normal benchmarks to test their model.