Average time is so misleading. it means that half of the requests were faster than this, but half were worse. but how much worse is not know
70% of the time I am in the i product I spend clicking on request and looking at the max number to see if it needs attention - by seeing that in any of the tables such as the "slowest" I could find and prioritize things so much faster.
90% is also an extremely useful metric - because I describe the time that most users experienced. the visual studio load testing software has this and the 95% percentile and its helpful, but does nothing for live systems with real humans