>>Are you sure that was 90% and not 95%? If it was 95% it probably refers to statistical / probability theory, which states that in a "normal distribution" 95% of occurrences will be within two standard deviations of the mean. In that case throwing out the outliers on the high and low ends of the distribution may give you a more accurate average. It's a similar concept to median average but not exactly, since with a median the high and low outliers offset each other in equal numbers. With the 95% method you are not as prone to extreme outliers on one side or the other which distort the average.
>>
>>I don't know offhand how to do this in SQL Server. You may be able to google up something up using keywords like sql server, average, normal distribution, and 95%. OTOH you might get about as good results using a median average. Why not give that a whirl first and see what your user(s) think?
>>
>
>Mike,
>
>It would be simple enough in T-SQL as I outlined in my prior message. Get the median value first and then get the sum of all values divided by the count of all values when the value is in 95% percentile.

The average of the middle 95% is not the same thing as the median.

Actually I'm not even sure what you mean by 95th percentile. Typically, for instance in test taking, the 95th percentile refers to the bottom 95%. I don't think that's what Kirk wants.