Douglas M. Hawkins

All of nature involves some unpredictable "random" variability, and manufacturing is no exception. It is inevitable, for example, that the diameters of the bolts holding auto engines together will vary, but it would be catastrophic if bolts too thin to hold the engine together were used in your car. So long as the variability is within set limits, the process making the bolts is said to be "in statistical control"; but when it exceeds these limits the process is called "out of control." Quality control (QC) involves two distinct methods for dealing with manufacturing problems: detection, where the object is to check whether the diameters of the bolts being used are within the acceptable tolerances; and prevention, where the object is to control the process so that the variability in the diameters never becomes unacceptably large. At the heart of the problem is the random variability from bolt to bolt. Statistical methods are essential for its resolution; and these methods constitute the discipline of statistical quality control. Although QC was born in the U.S. in the 1920s, American industry largely ignored it for decades. After World War II, American advisers introduced QC to Japanese industry, where it took hold and speeded the recovery of industry there. Today, American firms are adopting QC methods and are moving from the end of the process (for example, detecting batches of TVs with unacceptable levels of defects) back through the manufacturing operation (adjusting the machines making the TVs so that fewer of them have defects) back to the design phase (designing TVs so that they are inherently less prone to defects). All of these steps involve major statistical components. My own work in this area has been partly methodological and partly applied. The applied work has focused on improving the quality of plant operation, primarily in chemical industries. Methodologically, I have developed new ways of processing measurements to ascertain more reliably and quickly when they go "out of control." These new methods are starting to be used in operating plants, where they are expected to lead to lower variability and higher quality of production.