@dansup is bet a simple statistical approach could go a long way.

MAU/total user ratio would reveal a lot of the outliers I suspect. Especially if there are quick changes to this ratio over a day or week or so.