Reporting process capability: everything you always wanted to know many things you might not want to hear

There is a need to measure and report the ability to deliver high quality products and services to satisfy customers. Managers and engineers often resort to single number metrics for measuring the capability of their processes. Since the advent of Motorola´s ´Six Sigma´, there has been renewed interest in computing and reporting six such metrics: C_p, C_pk, C_{p tau} , C_pl C_pu and C_pm. However, these metrics are often misused by managers and engineers to ´beat´ their employees and suppliers into ´achieving high quality´. This leads to excessive inspections, rework and high costs of ´demonstrating quality´. This paper discusses methods for avoiding such misuse. Specifically, the author describes methods for computing each metric in the semiconductor environment where one must account for site-to-site, wafer-to-wafer and lot-to-lot variability. The author also computes wide confidence for each capability metric to illustrate the fallacy of using these metrics blindly.