The scope of this topic extends far beyond this meager post, by any account. Still, having put together thoughts in various posts about how to squeeze insights from benchmark data it behooves me to assemble, them starting at my 5,000th post (See my post of Feb. 10, 2010: visual analytics; Feb. 17, 2010: cost per hour and lawyers per staff; April 2, 2010: skewness describes data; April 26, 2010: a formula to show similarity between departments; April 30, 2010: two-group mean-comparison; May 18, 2010: whether to segregate data from very small departments; May 24, 2010: normalize spend data for a country by its tort costs; June 13, 2010: stem and leaf display of data; June 16, 2010: correlations between inside and outside spending figures; July 28, 2010: currency effect and the Big Mac index; Aug. 16, 2010: Kolmogorov-Smirnov two-sample test for representativeness; Aug. 16, 2010: total factor productivity and technology; and Nov. 10, 2010: normalized assets and revenues.).
There is even a metaposts in the lot (See my post of Sept. 30, 2010: normalized figures with 14 references.).