Have one benchmark number from the historical data and then compare the ad hoc result with this number to see if there is a huge variance. If there is a variance then check the process, if process ok then check the data source.