In my situation, we estimate cost and then compare the estimated and actual costs to get our estimation error: actual/original estimates.
Now, does it make sense to calculate an average estimation error and its SD? I believe weighted average is not required because the "error" is the same. The values, however, can be also negative (-100% means underestimated by 100%, while 100% means overestimated by as much). It would be a valuable information for us allowing to set certain tresholds once we have enough data, but I struggle a bit whether there isn't an issue I am overlooking.