Ordinary regression is the oldest and most popular statistical technique for predicting an outcome or dependent variable, such as sales dollars, salesunits, or any measure related to 'profit.' The adjective 'ordinary,' perhaps,suggests 'good old favorite' to the layperson. However, to the statistician, the word means something very specific. It refers to the method used to derive the regression model, namely, ordinary least-squares.

Ordinary least-squares should remind us that there is a key assumption of the regression technique, namely, the dependent variable data must be a 'normal' curve or bell-shaped. If the assumption is violated the resultant model may not be accurate and reliable.

Unfortunately, profit data from a solicitation is not bell-shaped. For example, a 2% response rate yields 98% nonresponders with profit values of zero dollars or some nominal cost associated with nonresponse. Data with a concentration of 98% of a single value cannot be spread out to form a bell-shape distribution. Accordingly, attempts at modelling profit with ordinary regression is questionable.

Fortunately, direct marketers can now explore the promise of a new method, the GenIQ Profit Model©, which maximizes profits from solicitations, without any concern of the shape of the dependent variable.