This paper deals with the development of a linear discriminant model for a large scale polypropylene production via the practical application of a multivariate statistical procedure. The objective is to differentiate between customer-acceptable and customer-rejection polypropylene product based on resin properties, metal residues, additive types and concentrations, moisture contents, and previous customer's responses. The discriminant model developed using this approach by including all significant product variables predicted correctly all of hopper cars shipped previously to customers for model validation. This statistical procedure was used as an interim quality control method for products sent to critical customers until a more definitive quality control procedure is developed.