I am working with some colleagues to set up alerts based on call center data, and I feel that something is wrong with how it is currently being done, but I'm not sure what exactly, or how to explain it.
We have hourly data on the amount of calls a center receives. We would like to set up alerts when it gets a low "amount of calls". We are exploring several distributions (like Poisson). If we have hourly data, and discover a lower bound, X, on that hour, does this mean that the lower bound for the minute is X/60?
Example: Let's say a site receives 40 calls per hour, and we find a lower bound of 5. Doing 5/60 = .0833 to get a minute bound doesn't seem to make much sense.
If we get 6 calls in an hour, then I don't think an alert should be raised, but if we scale the lower bound of 5 to a minute bound, then an alert could potentially be raised 54 times since those minutes don't have a call.