Featured Product
This Week in Quality Digest Live
Quality Insider Features
Richard Harpster
Good news? You are probably already doing it.
Adam Zewe
Researchers find the root cause of side-channel attacks that are easy to implement but difficult to detect
Constance Noonan Hadley
The time has come to check whether the benefits of teamwork still outweigh the costs
Lily Chen
The cornerstone of cybersecurity
Jeremy L. Boerger
To keep your business running, you need visibility into your IT assets

More Features

Quality Insider News
New standard leads to smoother production in 3D printing
Making designs a physical reality with the know-how to make more
Sapphire XC will ship in late Q3 beginning with aerospace companies
Major ERP projects take six months longer than companies were told
Program inspires leaders to consider systems perspective for continuous improvement and innovation
Collaboration produces online software for collecting quality inspection data
Serving the needs of employers and educators
Powder reuse schemes affect medical device performance

More News

Douglas C. Fair

Quality Insider

3 “Nevers” of Control Limits, Part 2

Never allow control limits to be automatically recalculated

Published: Monday, February 25, 2008 - 23:00

I was hiking in the Smoky Mountains National Park with a colleague, and we happened upon fellow hikers. Being the friendly sorts that we are, we stopped to take a breather and chat with our new female friends. We enjoyed resting and speaking with the women until my friend Bob smiled and asked one of them, “So when is your baby due?” The second woman’s eyes went wide, her hand covered her mouth and she slowly, carefully stepped away. Then, dead silence. The trickle of a nearby stream and the calls of birds miles away seemed deafening.

I looked at the woman who was the recipient of Bob’s query. Eyes had narrowed and jaws were clenched. She stared at Bob like a grizzly ready to attack. If you have ever made this mistake, you know “the stare.” Bob, still smiling and clueless, waited for a response. He never got one. After a few awkward moments, I grabbed him protectively by the elbow and tugged him down the trail path. “It was nice meeting you, but we really gotta go. See you down the trail!”

She wasn’t pregnant. And in case any of you aren’t sure, know this: Never ask a woman if she is pregnant. Not even if she’s in the obstetrics unit of a hospital looking like she’s hiding an oversized basketball under her gown.

Like ’ol Bob, there are lots of “nevers” in life to be learned. This is the second of three columns devoted to what must never be done with a control chart’s control limits. The first never was “Never Type In Control Limits.”

Ignoring the three nevers of control limits can place an entire statistical process control (SPC) deployment at risk of failure. Why? Because control charts and their limits help to identify when processes need adjusting, when they require problem solving, and, importantly, when process adjustments shouldn’t be made. Control limits must be deadly accurate, because we rely on them to tell us when to take action and when not. Effectively deployed, an SPC system can enhance a company’s bottom line, but if used improperly it can pummel a company’s bottom line.

For example, shutting down a manufacturing process because of a false statistical alarm could dramatically affect productivity and frustrate the people trying to determine what changed in the process. Alternatively, if an alarm is present but control limits are incorrect and don’t alert the operator, failure to fix the process could result in scrap, rework, and other costs that could negatively affect profitability and quality. Don’t forget the technological and human resource costs that are also required to support an ongoing SPC system. Collectively considered, all SPC system costs can be substantial, making it critically important that control charts and their associated limits are correctly calculated.

Because control limits tell us so much about how to manage and improve processes, you must pay close attention to them, and they must be correctly calculated. Following is the second never of control limits:

Never allow control limits to be automatically recalculated.

It might sound like a good feature, but if your SPC software allows control limits to be automatically recalculated at predetermined intervals, be forewarned: doing so will likely mask important process changes. We don’t want that, do we? Isn’t a control chart supposed to alert us to significant changes in data? We don’t want them to be ignored.

For example, say an operator is using an X-bar and range control chart. Imagine the process has a very gentle, gradual drift upward that is obvious only after viewing data over several weeks of time. Each day operators enter and view 25 subgroups of data. Control limits are recalculated every day. Each time the recalculations occur, control limits drift a little higher, the result of the slight upward movement in the data. Each day’s calculated control limits include the data’s gentle drift. Over short periods of time, the drift is imperceptible. Due to no fault of their own, operators could completely overlook the important process change solely due to the fact that the software automatically recalculates control limits.

So should control limits be recalculated? Yes, but only if there’s good reason to do so. In fact, control limits should be unchanging until these two statements are true:

      1. A statistically significant change has occurred in the mean or standard deviation.

      2. The change in process performance is expected to continue.

What about recalculating control limits after an engineering change or the inclusion of new machine tooling? How do you know if the new tooling has resulted in a sustained, significant change in mean or standard deviation? By plotting the new data against the old control limits, that’s how. If the new data show a different mean or standard deviation, and if we can expect that performance to continue in perpetuity, then control limits can be recalculated. The newly created control limits would be based on new data and would represent a new set of baseline control limits.

But that doesn’t always happen. Sometimes capital improvements don’t improve quality levels. Even if you’ve spent millions to improve a manufacturing process, new data should be plotted against old control limits. This helps to confirm that money spent actually had an effect on quality. It also helps to document process-improvement activities and quantify their results. Not until you can see a sustained, significantly different change in the data should control limits be recalculated.

You may be thinking that recalculating control limits requires a human to make the decision. If that’s what you’re thinking, you’re absolutely right. And that’s my point in this column: never automatically recalculate control limits based upon predefined times or subgroup intervals. Doing so can mask critical process-control information that could put your SPC system at risk.

Accurate, correctly calculated control limits can help manufacturers reduce costs, improve quality, and increase customer retention. Through their proper use, control limits can mean a big difference in terms of product quality and costs. Control limits that are automatically recalculated, however, can have seriously negative effects on an SPC deployment. So the next time someone says to you, “hey, I want to automatically recalculate the control limits.” Just give them the stare that Bob knows so well. You know the one. It’s the one that implies, “You shouldn’t even be asking that question.”

Discuss

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).