So the metrics where misleading because testing was not constant and/or performed same way across the same population distribution?