Monitors like CPU, HTTP Response time, Memory etc show you an average in the graphs.
We currently alert when a poll (or 2 or more consecutive polls) go above the alert level, But we would like to alert if the average is higher than "normal".
i.e. out HTTP monitors alert when they hit 500MS or more, they usually run at between 10-100 with the odd spike. However the average is usually around 30MS. We had one monitor not alert but the average hit around 200MS for a week.
So it would be good to alert if the average for the last 12 hours is higher than the average for the last week for example.