As I commented on a call, the mean is not a useful measure in this situation. The median is the 50th percentile, which is useful. And instead of -- or perhaps in addition to -- the histogram, I suggest a handful of other percentile measures: 10, 30, 50, 70, 90. These will give a pretty good picture of the distribution of processing times. An alternative approach would be to choose a handful of ranges, e.g. - Under 5 seconds - 5 to 60 seconds - 1 to 10 minutes - 10 minutes to one day - one to three days - three days to a week - Over a week The ranges should be based on a combination of understanding of what might cause variations in response time, e.g. whether the request is subject to manual review vs automatic processing, and which thresholds are relevant to the requester community. If we don't have a good sense of where to set these thresholds -- and I presume we don't yet -- then the percentile breakdown above will be helpful in understanding the variations of response times and point the way toward further discussion and perhaps further analysis. I edited the document to reflect my comments in the first paragraph. Thanks, Steve