Using a P Chart for Quality Control

Control charts incorporate statistical control limits that show the amount of variation in process performance due to common cause variation, meaning it is inherent in the process itself. If any points lie outside those control limits, the proper conclusion is that special cause is present. Other special cause tests provide information about the existence of trends or process shifts.

Without control charts, business leaders often misinterpret data, leading to actions that either do not address the underlying problem or, even worse, exacerbate the problem. For instance, it is tempting to assume that if a metric increased from one month to the next, something specific must have changed resulting in different process performance. However, random variation exists in any process, and it is possible that nothing at all changed, and that the following month the metric will decrease again.

This type of mistake is particularly problematic when managers are conducting employee performance evaluations or running an incentive program. For instance, if one customer service representative has an increased average call handle time, it may in fact be due to a slip in efficiency, but it could also be because of variation in the types of calls all representatives handle.

Because control charts provide information about the variation inherent in a process, they give leaders a good idea of the range of performance they can expect from a process in the future. By comparing this range to the customer specification or business requirements, they can determine whether the process is capable of achieving the target level of performance or whether process improvements are needed. Learn more about control charts in Introduction to Control Charts and Types of Control Charts.

Use a p chart for quality control initiatives when your data is of the discrete attribute type, also known as categorical data. In other words, if your data is obtained by grouping instances into specific categories, you have discrete attribute data. For more information about types of data, check out our article, Six Sigma Data Types.

The p chart plots the proportion of data in a specific category. You do not need to have equal sample sizes for each data point, but if you do you can instead use a variation of the p chart called an np chart. Typically the proportion data is plotted over time, so that business leaders can determine whether process performance is stable or shifting. For instance, a Six Sigma project team might use a p chart to assess whether the proportion of loan applications approved is decreasing, or whether the proportion of software support calls about program installation is rising. Using a program like Minitab they can identify process shifts and trends, based on the data distribution. Teams conducting DMAIC projects often use a control chart to depict a change in process performance before and after improvements are implemented.

P charts can also be used to analyze data that is arranged in different groups rather than plotted over time. For instance, rather than assessing whether the loan application rate is changing over time, a financial company's leaders can compare application rates for different types of loans, different customer segments, or different processing teams. Because the sample size may differ for each group, the control limits can actually vary for each data point. Each point is thus compared to its own limits to identify outliers, with any point above its upper control limit indicating that group measures higher on that metric than the others, and any point below its own lower control limit representing performance below the other groups.