The difficulty is in the details. Normalizing would have to account for a variety of different variables such as location, average age/customer base, average revenue per customer, etc. Furthermore, you would need all of that data many different businesses to even be able to build your normalization.

Sure, you could do a subpar job with the normalization but then that will increase your false positives and decrease confidence in the method.