"The
needed data cannot, in most cases, be culled from existing human resources data,"
he says. "To create the needed data and analysis, human resources executives
must run experiments within their organizations. They must invest in efforts to
train departments in group processes, and then follow their performance over
time, comparing the performance of groups that have been trained with that of
groups that have not, using hard performance measurements based on the goals of
the unit." These measurements might be time-to-market, error rates, sales
growth, or any number of other productivity measures.