Logistic regression is used to deduce the relative importance of various drivers in order for the dependency variable to be recreated when the dependency variable is binary (for example, yes/no or buy/not buy). It is used when the usual linear regression is not possible and is especially useful in the modeling of propensity.