Your process should be stable and the original (or transformed) process data should follow a normal distribution. The probability plots and the fitted normal curves allow you to check for potential problems.
Use the normal probability plots to assess the requirement that your data follow a normal distribution.
If the normal distribution is a good fit for the data, the points form an approximately straight line and fall along the fitted line that is located between the confidence bounds. Departures from this straight line indicate departures from normality. If the p-value is greater than 0.05, you can assume that the data follow the normal distribution. You can evaluate the capability of your process using a normal distribution.
For each variable, compare the solid overall curve and the dashed within curve in the histogram to see how closely the curves are aligned. A substantial difference between the curves may indicate that the process is not stable or that there is a significant amount of variation between subgroups for that variable. Use a control chart to assess whether your process is stable for the variable before you perform a capability analysis.
If your process naturally exhibits large variation between subgroups, such as a batch process, and the variation is not due to special causes, choose the Between/Within subgroups option when you perform the normal capability analysis for multiple variables. If you use the B/W analysis, a substantial difference between the two curves may indicate a systemic source of variation in the process, in addition to the variation between and within the subgroups.
For each group or variable in your data, use the capability histogram to visually examine the sample observations in relation to the process requirements.
For each variable, visually examine the data in the histogram in relation to the lower and upper specification limits. Ideally, the spread of the data is narrower than the specification spread, and all the data are inside the specification limits. Data that are outside the specification limits represent nonconforming items.
To determine the actual number of nonconforming items in your process, use the results for PPM < LSL, PPM > USL, and PPM Total. For more information, go to All statistics and graphs.
For each variable, evaluate whether the process is centered between the specification limits or at the target value, if you have one. The center of the data occurs at the peak of the distribution curve and is estimated by the sample mean.
Use the main capability indices to evaluate how well your process meets requirements.
Use Cpk to evaluate the potential capability of your process based on both the process location and the process spread. Potential capability indicates the capability that could be achieved if process shifts and drifts were eliminated.
Generally, higher Cpk values indicate a more capable process. Lower Cpk values indicate that your process may need improvement.
Compare Cpk with a benchmark that represents the minimum value that is acceptable for your process. Many industries use a benchmark value of 1.33. If Cpk is lower than your benchmark, consider ways to improve your process, such as reducing its variation or shifting its location.
Compare Cp and Cpk. If Cp and Cpk are approximately equal, then the process is centered between the specification limits. If Cp and Cpk differ, then the process is not centered.
Use Ppk to evaluate the overall capability of your process based on both the process location and the process spread. Overall capability indicates the actual performance of your process that your customer experiences over time.
Generally, higher Ppk values indicate a more capable process. Lower Ppk values indicate that your process may need improvement.
Compare Ppk to a benchmark value that represents the minimum value that is acceptable for your process. Many industries use a benchmark value of 1.33. If Ppk is lower than your benchmark, consider ways to improve your process.
Compare Pp and Ppk. If Pp and Ppk are approximately equal, then the process is centered between the specification limits. If Pp and Ppk differ, then the process is not centered.
Compare Ppk and Cpk. When a process is in statistical control, Ppk and Cpk are approximately equal. The difference between Ppk and Cpk represents the improvement in process capability that you could expect if shifts and drifts in the process were eliminated.
The Cpk and Ppk indices measure the capability of the process in relation only to the specification limit that is closest to the process mean. Therefore, these indices represent only one side of the process curve, and do not measure how the process performs on the other side of the process curve. If your process produces nonconforming items that fall outside both the lower and upper specification limits, use additional capability measures in the output to more fully assess process performance. For more information, go to All statistics and graphs.