Within the possible linear relationships that can exist in a multiple linear regression model (multicollinearity), one relationship that is often overlooked is the one between the constant term of the model and the rest of the independent variables. This type of approximate multicollinearity is ignored by the variance inflation factor while it can be detected through the condition number or the coefficient of variation. In this work, the utility of the metric number is analyzed for detecting this type of approximate multicollinearity from a geometric perspective.