Control limits and specification limits: Where do they come from and what are they?
After talking to customers when I worked in Technical Support and more recently at JMP Discovery Summit conferences and site visits, I realized that there is confusion about the difference between control limits and specification limits. While you may have heard of, or even used both of these in your work, they are quite different from each other. In this blog post, I will explain those differences and hint at how you can use each of these limits in JMP. Future blog posts will expand on these ideas.
Specification limits come from your customer. They are limits on product characteristics that define where the product works and where it does not work. Spec (Specification) limits define the allowable deviation from a target or ideal value.
Sometimes these specs are engineering specs. If the product falls outside of these specs, the item is rejected or will not work. For example, company XYZ manufactures nuts. These nuts are supposed to fit onto a specifically sized screw, say a ¼” machine screw. If the nut is too big, then it will not hold the screw in place. If the nut is too small, it will not fit on the screw at all.
Spec limits should be placed at the point where losses due to variation equal the benefit of the product. These limits are often, but do not always have to be, symmetric. The customer may want to err on the higher side or the lower side depending on the application.
Spec limits are used in capability analysis, which enables you to assess if a system is statistically able to meet a set of spec limits or requirements. Capability analysis helps you to determine if your system is capable or incapable. JMP offers capability via several menu items. You can find this analysis in Process Capability, Process Screening, Control Chart Builder, and Distribution.
Control limits are based on the performance of your process. These values are calculated from data, and they tell you about the variability in your process. Control charts are used to calculate these values. An upper control limit (UCL), center line and a lower control limit (LCL) will be calculated. Generally, the limits are placed at center line +/- 3*sigma where sigma and center line are calculated based on the type of control chart chosen. The intent is to use these calculated control limits to tell you when your process has changed and adjustments need to be made. JMP offers a wide variety of control charts to choose from (Shewhart, Attribute, Rare Event, UWMA, EWMA, CUSUM, Presummarized, 3-Way, Levey Jennings, and Multivariate Control Charts).
The control limits calculated via the different types of control charts allow you to determine if your process is stable, or in control.
Control Limits vs. Specification Limits
Calculated from data
Given by customer or design
Based on variability
Based on system requirements
Applied to summary statistics
Applied to individual measurements
Applied to process measurements, perhaps on product, perhaps not
Applied to product characteristics
The voice of the process
The voice of the customer
In the Process Performance Graph below, note that your process can fall into four different areas. The green area denotes processes that are both capable (based on spec limits and capability analysis) and stable (based on control limits and control charts). About 1/3 of these processes are both capable and stable. The yellow area denotes processes that are capable, but unstable. None of these processes fall into that category. The orange area denotes processes that are Incapable, but stable. About 2/3 of the processes fall into this category. The final category is red, which indicates incapable and unstable, which none of these processes fall into. In this graph, all of the processes are stable (in control), but only 1/3 of them are capable (meet the specification limits).The Process Performance Graph divides the processes into the four categories of capability and stability.
I hope this post has helped clear up the difference between control limits and specification limits. In the future, look for more posts that expand on this topic and go into more detail about Capability Analysis and Control Charts.