We outline our statistical modeling strategy to combine responses to multiple questions on a client satisfaction survey into a single measure. We begin by describing the business problem that drove this analysis. Then we describe our strategies for imputing the missing data in the survey questions, reducing the variable set, specifying and transforming the dependent variable, and assessing impact of multicollinearity as we develop a statistical model. We explain the interplay of multivariate statistical techniques - principal component analysis, classification trees, and logistic regression - to gain different perspectives of the data. And how we finally succeeded in developing a single measure of client satisfaction.