Survey analysts ought to explore correlations among data, and quantify the relationships

Too many “analyses” of survey data do no more than regurgitate the findings. “Twenty percent of all the law departments have receptionists.” That style of factoid reporting has a bit of value, but an analysis that matches the data against another set of meaningful data is far more useful. That process is called correlation (See my posts of April 5, 2005 and May 10, 2005 on correlation.). “The more they spend on receptionists, the higher a law department’s retention rate.” Even better than a statement of linkage is a finding such as “For every $10,000 spent on receptionists, a law department increases its retention rate by 10 percent.)

Another example might start with a bare metric: “The average law department in this survey had three locations.” A linkage correlation would be along the lines of, “Law departments with more offices tend to have lower inside costs per billion dollars of revenue.” A specific quantification would show that “Every three additional locations drops inside legal costs per unit of revenue by eleven percent.”