Some of the issues faced by institutions as they attempt to design a system of constructs that reflects the diversity of their schools are addressed. The choice of statistic, percentage passing versus mean, used in reporting test results can impact the allocation of services to subpopulations in the school. Percent passing statistics tend to focus resources on students whose scores fall near the cutoff for passing, while annual comparisons of mean scores tend to be affected most by students scoring at the extremes of the distribution. Sample bias is another issue that cannot be ignored. Test scores can be affected by large numbers of geographical transients. Errors of omission can have a snowball effect on statistics such as attendance, and distortion attributable to omission can affect test results as well. Administrators must look beyond common sense indicators of school success to construct statistical profiles that reflect disparate populations fairly. Indicators must be designed so that community pressure to show improvement does not, in fact, reward or ignore deleterious practices. (SLD)