I'm not a fan of averages. Truth be told I think averages are evil.
I normally go into an "average" rant whenever a new developer starts doing performance testing. Said developer will then give me "average" times, thinking it indicatesa program/process' performance. When I explain to said developer why averages (or they way he calculated) are useless, he (normally they) still don't believe me. (Most of the times though it's probably because their ego can't handle it)
Recently I experienced it again. (Although, to this developers credit, he understood the problem as soon as I explained it.)
The problem again had to do with performance testing. In our application, we need to do calculations on various products and, understandably, the various calculations were different. The developer then came back with the following results (timings given are ficticious but indicative)
Prod1: 0.01 ms
Prod2: 0.01 ms
Prod3: 0.05 ms
Prod4: 0.70 ms
Average time 0.19 ms
Now this average looks acceptable (as a response time), except that most of our users will actually choose Prod4. I.e. their "average" is actually 0.7 ms and NOT 0.17 ms. Almost half a second difference.
So, what am I trying to get accross. Make sure your average is really the average you think it is. ;)
Greets
Roaan