We are tracking a set of monthly business metrics (some are absolute numbers and some are %). For each metric, we want to find a range of variations that tells us if the current month's performance is as expected or warannts a callout. We are expecting that if we are not deploying a major change in the technology, a metric should fall in the range. There might be some seasonality to the data. Ranges should be given relevant to most recent month.
My first idea is to analyse the data, see if they follow some kind of distribution and use the standard deviation to create the band. Another approach could be to remove the outliers and look at a moving average of last x months.
I am not sure how to capture the seasonality of data, e.g. we are expecting much higher variations in holiday season.
What would be your suggestions to approach the problem and if you can recommend a resource to read more about the topic?
Thanks in advance!