Featured Product
This Week in Quality Digest Live
Quality Insider Features
Rita Men
A survey shows people tend to trust their employers more than governments or the media
Jo Napolitano
Scientists developed a device that can sort information similarly to how the human brain does
Optical Gaging Products OGP
OGP Starlight 150 vision system helps watchmaker measure small, complex watch parts without operator interpretation
Knowledge at Wharton
A hybrid workforce will be much more difficult to manage
Del Williams
For static-sensitive powders in certain concentrations, controlling potential dust ignition, escape, and buildup is critical

More Features

Quality Insider News
Tech aggravation can lead to issues with employee engagement, customer experience, and business results
Reliance NXG: Cloud-native and analytics-driven solution is a catalyst for innovation and digital transformation
Four-axis models enable a wide range of palletizing applications, order picking, and other logistical tasks
Two new standards join the ISO series
Location-independent real-time monitoring, predictive maintenance even for purely mechanical pumps
Online software for collecting measurements, visual-defect information, and pass/fail situations
ProMation announces additional options for constructing motor-operated valves for industrial flow control
Simplifies workflows and enhances ease of use for manufacturing, assembly, and construction
Harnessing the forces that drive your organizations success

More News

Douglas C. Fair

Quality Insider

3 Nevers of Control Limits, Part 3

Never arbitrarily widen control limits

Published: Monday, April 28, 2008 - 22:00

Never say “never”? I guess I overlooked that memo. In case you missed it, the last two columns were written about what you should never do with control limits. These three nevers, if avoided, will ensure that your control charts are useful, reliable tools.

The first two nevers of control limits are:


Never allow control limits to be typed in.


Never allow control limits to be automatically recalculated.

If any of the nevers is undertaken, you could be left with a control chart that doesn’t identify important process changes. But isn’t that the whole idea? Shouldn’t control limits be set so that they alert users to unusual circumstances? Well, yes. That’s the idea, and that’s why control charts are so extraordinarily useful.

Lots of different alarm rules can be applied on a control chart. But this column is written to specifically address the most commonly applied Western Electric rule of all: points that fall outside of control limits.

In my experience, I have unfortunately found that a small proportion of SPC users find control charts to be annoying. They’re a necessary evil and the less time and energy expended with them, the better. These people think plot points should always fall within control limits, and if they don’t, they become mightily unhappy. They become irritated because someone must take precious time away from what they’re doing to respond to control chart alarms.

Because this minority of SPC users doesn’t want to waste time reacting to statistical events, they sometimes try to alter control limits to minimize the possibility of alarms. This is not just a bad idea; it’s dead wrong. Hence, the third of the three nevers of control limits:

3. Never arbitrarily widen control limits

SPC coordinators may recognize this never as requested in this way: “Hey, Bob, I want you to set the control limits at ± 6 standard deviations instead of ± 3 standard deviations. Can you do that for me?”

In most cases, these requests basically boil down to this: “I want my control limits to be so wide that it will be virtually impossible for plot points to fall outside of them.” If limits are set artificially wide, then there exists little or no chance of those pesky out-of-control limit alarms being triggered.

The more insidious interpretation of the request above is: “I want to hide the fact that my process is out of control.” As discussed in prior columns, a control chart that signals an out-of-control event shouldn’t be viewed negatively. It’s a good thing that a tool can alert us to potential process problems. That information could be critical for heading off some previously unforeseen issue that could cost a manufacturing organization a huge amount of money.

Let’s go back to the basics
Assume that a process is normally distributed and that the mean and standard deviation are unchanging. If control limits are set, as Walter Shewhart, Ph.D., specified, at ± 3 standard deviations, then 99.73 percent of all normal data values should fall between those limits. If plot points fall within control limits, then it is an indication that the process is operating normally. Although normal may not equate to acceptable, control charts with plot points within limits indicate a consistent, normal operation. When a plot point exceeds control limits, there is only a 0.27 percent chance of being a normal event. Therefore, because of this miniscule probability, we presume that something abnormal has occurred and that the process needs attention.

If control limits are widened significantly, then the possibility of detecting an abnormal event decreases dramatically. In essence, unless a plotted data value is wildly different than previously, no out-of-control limit alarms will be triggered. No one will be alerted to important changes in the process. No one will be aware of critical issues that could adversely affect product quality.

By arbitrarily widening control limits, users eliminate control chart usefulness and undermine the effectiveness of expensive, vitally important quality systems. If a control chart is triggering alarms, then it’s doing its job. It shouldn’t be viewed as bothersome or annoying. Most SPC users are extremely grateful when their control chart indicates a process change. These experienced SPC users know that their control chart’s alarm is helping to sidestep potentially serious problems in the future.

We must be very careful with the way that control limits are calculated and how they’re applied. They must be accurate to be useful. They must not be arbitrarily widened, typed in, or automatically recalculated. If any of these nevers is undertaken, control charts could spit out erroneous alarms, pushing an organization into ill-advised and unnecessary process changes. Likewise, if control charts never trigger alarms, reasonable people might ask how SPC is useful and why money is spent on it. Abiding by the three nevers will allow you to avoid these unnecessary errors and provide confident, reliable indicators of required action or, conversely, inaction. And for those who have advised me to never say “never,” well, I’d never do that.


About The Author

Douglas C. Fair’s picture

Douglas C. Fair

A quality professional with 30 years’ experience in manufacturing, analytics, and statistical applications, Douglas C. Fair serves as chief operating officer for InfinityQS. Fair’s career began at Boeing Aerospace, and he worked as a quality systems consultant before joining InfinityQS in 1997. Fair earned a bachelor’s degree in industrial statistics from the University of Tennessee, and a Six Sigma Black Belt from the University of Wisconsin. He’s a regular contributor to various quality magazines and has co-authored two books on industrial statistics: Innovative Control Charting (ASQ Quality Press, 1998), and Quality Management in Health Care (Jones and Bartlett Publishing, 2004).