“Proving quality” may seem to be an elusive concept: what can it mean to customers and suppliers? After all, just insisting that your products are of high quality isn’t enough for any of the stakeholders, and yet many organizations adopt this approach anyway. Marketing language won’t mask underlying challenges to predictable, dependable quality.
Just as there are tools to support the improvement process, there are tools to demonstrate once and for all that your organizational practices assure the quality of your products. Among these tools are
- Process behavior, Shewhart, or control charts
- Statistical process control
- Measurement systems analysis
- Gage calibration management
- Process capability analysis
No single tool or approach is sufficient in the ongoing need to communicate quality. Instead, you need a panoply of approaches that address your measurement systems, your processes, and the capability of those processes.
One critical part of this communication is the visual impact of control charts that show the behavior of a process, and the ensuing use of statistical process control (SPC). Data alone—in list form, perhaps—does not have the power that a chart carries. This may be because of the chart’s ability to clearly show trends or out-of-control conditions, or because of the inherent analysis of the data that a chart conveys. Data facts are fine, but have limited usefulness without appropriate analysis. Making meaning from statistics is quite different from simply gathering data.
Here are some football statistics, for example, that lack meaning: 14-0; 7-10; 36-28; and 18-14. To create meaning, further steps are demanded, answering critical questions: Who was playing? and Which team had the higher score? These basic questions must be addressed in order to begin to analyze the outcomes. Otherwise, the data has no meaning. For SPC to yield information, essential questions about process stability must also be addressed, and the analysis must be supported by identifying markers of instability. As we will see, these markers are straightforward indicators of process control.
Here is a column of data related to the viscosity of oil:
And what follows is the same data but charted on a control chart.
Statistician Walter Shewhart, known as the father (or grandfather) of statistical process control, saw the need to distinguish between “common cause” or “assignable cause” variation, and “special cause” or what was once called “chance cause.” Every process demonstrates variation, but addressing causes of special variation is the first step in reducing that variation. Reducing common cause variation demonstrates that the process is becoming more consistent. Control charts are interpreted by using basic rules to assure that processes are in fact stable (in control). These rules include identifying instability when certain conditions appear:
- Data points that fall outside control limits;
- Seven or more points in a row that fall above or below the center line;
- Seven or more points in a row that go in one direction (up or down);
- Presence of nonrandom patterns.
(While these represent commonly used rules, other guidelines exist. Some organizations use eight points, rather than seven, while others may use six.)
It is clear that these rules cannot easily be applied to raw data alone, but instead the data must be reflected on a chart that will clearly demonstrate trends such as these. This is where the analysis comes in. Further, the data itself will determine where control limits fall, applying statistical formulas by hand or automatically with software.
Showing that processes are stable is perhaps the most dramatic way to demonstrate to customers, suppliers, and other stakeholders that your organization’s products and services are of predictable quality. Charts that reflect stable systems are proof of your quality.
Next month, we will look at the role that measurement systems analysis, gage calibration management, and capability analysis contribute to proof of quality.