I am working on using prediction for alerting. I want to track some data I have that has some business trends; it is fairly seasonal. Right now we manually calculate a baseline; which only works if our data is stable. So I have found as our traffic rises and falls through the seasonality of it; our calculated baseline is inaccurate most of the time. I have created a prediction based alert but it has not been received too well since it uses standard deviation. Is there an issue to base alerting from standard deviation? Just trying to under stand if there is some limitation or if it is error prone. The data I am getting seems to be accurate thresholds for the alerts.
