# Statistical training for quality improvement.

SPC has become part of most manufacturing processes and its concepts can be introduced on the shop floor in a short time

There are liars, damned liars and statisticians." Benjamin Disraeli "There are three kinds of lies: lies, damned lies and statistics." Mark Twain.

People have an inherent mistrust of statistics.

Our background makes us wary of these simple techniques that take away pages and pages of hard numbers and give, in return, nothing more than a mean and a distribution. Any successful statistical training program must deal with a dual role; it must both teach the concepts and, at the same time, overcome the mental blocks that have been erected due to the suspicion against statistics and their misuse.

Data and data management

Most process monitoring schemes result in the acquisition of pages and pages of hard numbers. Usually, these numbers are given a quick check, to see that they are in compliance with the operating specifications, then filed. They will sit in the file until some troubleshooting or optimization effort requires that they be retrieved. The central limit theorem tells us that, in the long run, that collection of data can be reduced to a mean value and a spread or distribution.

For measured data, that distribution will generally be a normal or Gaussian distribution. It has a well defined mathematical form and represents the type of spread that can be expected in many systems if the plant is well-controlled. The data will contain a small proportion of low and high values. These are real and part of normal operation and it is this ability to think statistically that provides the methods needed to distinguish this randomness from a process upset.

Statistical thinking was first applied in a systematic fashion to plant process data by Shewhart at Western Electric in 1924. The new fields of statistical process control (SPC) and statistical quality control (SQC) were revolutionary to industry and brought a new rationalism into data collection.

Individual points combine to form distributions. It is the behavior of the distributions that describe the long-term behavior of any controlled system, not the individual points. The time had come to realize that individual points are not sacrosanct.

With the adoption of SPC, the role of data collection has changed to one that is interactive with the process. The new goal is to optimize control. Today SPC has become part of most manufacturing. It is a mandatory requirement for all suppliers to the automotive industry.

The same type of analysis can be applied on a variety of systems. Often the results of a rational analysis can be quite informative ... and quite different from the accepted concepts held by people, who may be looking at them and unconsciously seeing what they want to see.

The original Western Electric work was one of those ideal situations where the statisticians recognized the practical aspects of an operating facility. There were no computers or calculators. In keeping with that original concept, a training program designed to teach statistical thinking does not go deeply into statistical theories. If SPC is truly for the use of the worker on the shop floor or the operator in the control room, the concepts must be simple enough that they can be incorporated into everybody's day-to-day modus operandi.

The training outlines a few simple statistical principles then goes out to apply them. It takes roughly a day to introduce the concepts of SPC and the background necessary to start applying them. From the instructor's point of view, it can be a real thrill to watch scepticism turn into acceptance.

Company wide quality improvement

Improving the quality of the data is the first of many steps needed for a successful Company-Wide Quality improvement (CWQ) program for the process industries. The ultimate objective is to improve the capability of the process. (Capability in SPC terminology is the measure of the statistical distribution of the process output compared to the target.) To do so calls for the development of an entirely new way of thinking, the heart of which is a strategy we like to call Management Using Statistical Thinking, or MUST.

The goal of the training program is to provide all levels within a company, from technical specialists to senior management, with a statistical-knowledge toolbox. The tools use relatively simple statistical principles to gain a deeper understanding of the processes involved and provide a variety of techniques that aid in troubleshooting, setting priorities, and establishing operating standards.

Education in statistical principles and training for the basic tools, such as capability measurement, Pareto analysis and value analysis, is best done in mixed groups of managers, specialists and operating people. Training in the use of the larger set of analytical tools, including analysis of variance, design of experiments, cause-and-effect analysis and quality function deployment, is confined to a small group of technical experts. For any CWQ program to succeed, it is vital that senior management be committed to the program and that they be trained along with the workers.

One of the best ways to get a CWQ program operating, has been through the formation of venture teams drawn from a variety of disciplines or work areas.

In addition to the ability to use the statistical tools, team members need to be taught the methodology for continuously improving quality and how to work together as teams. It is the team members who would be the first to apply new tools.

They have to be prepared to encounter the inevitable clash between Culture and Technology. In any industrial environment, the CWQ concept would tend to be dropped if it doesn't show success within a relatively short period. Without proper training and coaching, the process could be long and chancy. With the proper training, some initial successes are more likely. These will prompt the team feeling to emerge and real progress can be possible.

The adoption of a CWQ program is a commitment for an ongoing retraining program. Each new step produces an increase in our understanding of the plant processes and how they operate. It is this ever-growing knowledge base that enables us to bring about continuous quality improvements.