A method for analyzing the characteristics of purchasing quality costs using a probabilistic model with random input parameters
This dissertation develops a technique for predicting purchasing quality costs and studying their behavior as random variables. The resulting model extends prior research which treated purchasing costs deterministically and it thereby provides management with a valuable decision and planning tool. This is important because the trend is for companies to place an increasing reliance on purchased components.
Expressions for the expected cost of three alternative policies are developed: incoming evaluation of components, evaluation of components at the supplier, and waiving of product evaluation with a review of supplier data only. The cost of off-site evaluation (at the supplier), as an alternative to incoming evaluation, has not been considered in previous literature. Important quality cost inputs are treated as random variables with frequency distributions estimated from company historical data.
In critical applications where safety of the end user is the primary concern, it is desirable to detect faulty components at the earliest possible stage in the production flow, and some form of product inspection will always be desirable. The model proposed in this dissertation is particularly suitable under these conditions.
The model enables the management practitioner to justify switching between policies based on expected costs. Cost data from a small U.S. Government contractor demonstrates the usefulness and application of the model. A computer simulation is used to illustrate the application of the model and a method for developing probability distribution functions from empirical data is discussed and demonstrated. The simulation results suggest that an evaluation policy based on management consensus or intuition may not always be the lowest cost policy.