February 27, 2007 |

TRENDING Root Causes, Accidents, and Other Infrequently Occurring Statistics


“Figures often beguile me,
particularly when I have the arranging of them myself;
in which case the remark attributed to Disraeli
would often apply with justice and force:
‘There are three kinds of lies:  lies, damned lies, and statistics.’”
Mark Twain autobiography, 1904


While the individual man is an insoluble puzzle,
in the aggregate he becomes a mathematical certainty.
You can, for example, never foretell what any one man will be up to,
but you can say with precision what an average number will be up to.
Individuals vary, but percentages remain constant. So says the statistician.
Arthur Conan Doyle

An except for the upcoming revision of the TapRooT® Book…
(Copyright © 2007)


Many people think they can spot a trend by looking at a bar or line graph. They can’t. Dr. Walter Shewhart’s work at Bell Labs proved that people can’t “eyeball” trends (Economic Control of Quality of Manufactured Product, D. Van Nostrand Company, 1931). If you try to manage by eyeballing trends, you actually make the process worse. This is the first way that trending gets a bad name.

How should trending be used? You should use statistics properly to measure performance, select targets for improvement, and detect significant trends.

How should you target areas for improvement? By using numerical measures of performance based on statistical analysis of either reactive or proactive data based on the measures previously described in this chapter.

How should you measure performance changes and detect significant trends before major accidents, quality problems, or plant upsets occur? By using methods developed for statistical process control to determine which trends are part of normal process variation and which trends are the result of actual significant changes in the system.

Thus, trending goes far beyond common practice of looking at a bar graph and saying, “up is good and down is bad.” This new way to trend can help you prove that performance has significantly improved or taken a turn for the worse. These graphs aren’t just someone’s opinion. This “new” way to trend is based on rules for optimizing performance that have been proven by Shewhart since the 1920’s.

Trending can:

•Use reactive measures to show the actual impact and payback of your improvement programs.
•Use proactive measures of key performance elements to predict where problems might occur so that you can take action before the problems cause major production outages, quality issues, or accidents.
•Use reactive data and show the areas where you will get the most “bang for your buck” when you spend time, money, and effort to improve performance.
•Show you that there aren’t any areas that will provide amazing returns for the effort invested.

You may ask, “If these statistical rules are so good, why haven’t I heard of them and why isn’t my company using them?” The answer is twofold:

1. Most mathematicians don’t understand the simplicity of this trending. They make things much too complex. Too hard. They get lost in p-charts, t-tests, and normal distributions. And they adopt formulas that just don’t work in the real world to predict performance.
2. You may have heard of the techniques but, once again, they seemed too hard. The techniques shared in this chapter are the basis for Statistical Process Control and Six Sigma.

There is one more way that trending gets a bad name. Managers, without statistical training, get frustrated with the unintelligible voodoo analysis of mathematicians. They can’t understand the basis for the trends and they just don’t want to waste their time. So what do they do? They “simplify.” How simple? How about up is good and down is bad. Then every reporting period people scurry around trying to find reasons to justify the normal variation that every process experiences. And we are right back to the bad practices that are proven NOT to work that we started this section discussing.

What if you don’t believe me? Let’s look at an “example” trend…

(To see the graphs – you’ll have to buy the book – they wouldn’t post here)

The figure above shows a graph of safety related incidents per month. Since safety related incidents are bad, up is bad on this graph. And if we used the normal straight-line approximation to “trend” and predict performance, we are at a crisis stage.

What should we do? Hire consultants. Fire the Safety Manager. Conduct a one-day safety stand down. Have the CEO make a video where he emphasizes that safety is a top priority. But wait … what was last month’s top priority? Oh, don’t bother me with details! This is a crisis! We need improvement and we need it fast before something really bad happens. Why? Because safety is out of control! Anybody can see from this graph that bad things are going up at an alarming rate. We must act now and act decisively!

And that’s the way trends are interpreted at thousands of companies and facilities around the world. Unfortunately, the answers we just interpreted from the graph by using a straight-line approximation are wrong. Wrong? How can they be wrong? Look at the figure below. The four points from the figure above are the same as the first four points in the figure below. No real action was being targeted to change safety performance. So the variation we are observing is just part of the random variation that happens all the time. There really was NOT a crisis.

(To see the graphs – you’ll have to buy the book – they wouldn’t post here)

What is wrong with reacting as if there was a crisis when there isn’t one? The attention to safety doesn’t hurt does it? Yes it does! Every child knows the story of the boy who cried: “Wolf!” and the story of Chicken Little crying: “The sky is falling! The sky is falling!”

Look at the results in the false trend example.

1. The Safety Manager lost his job unnecessarily. (Certainly not a positive outcome for the Safety Manager.) The new Safety Manager is more likely to overreact to future trends and cause even more unnecessary emergency changes.
2. Money was wasted hiring consultants and implementing fixes that probably were a waste of time. Since time, money, and effort are scarce resources, they had to be taken away from some other work. This other work (that may have been necessary) was not completed on time. Therefore, an opportunity was lost.
3. Employees become jaded. Why? Because every month (day or week) there’s a new crisis. A new set of priorities. A new video from the CEO. Sooner or later, people lose their sense of urgency. They lose faith in their management. They stop reacting to new initiatives because they know that these will blow over just like the last management fad. They become complacent.

You might think that this is bad. But we’ve seen worse. Much worse. We’ve seen millions of dollars wasted on improvement programs that were scrapped just months after they were completed with NO payback on the investment. We’ve seen hundreds of people transferred or “right-sized” during a crisis transformation program in reaction to imagined trends.

Finally, we’ve seen people stop reacting. Why? Too many crises. People who cried wolf too often. And there wasn’t a way to clearly communicate the lessons that could have been learned from the real trends. A real crisis was at hand but nobody reacted appropriately. They were simply overwhelmed by false crises and didn’t react to the real crisis.

We believe that this was one cause of the BP Texas City Explosion and Fire. There was a data warning management of a process safety problem, but it wasn’t analyzed and presented appropriately. The result? The loss of 15 lives. Turmoil in management. Increased gasoline prices due to even tighter refinery capacity. These fatalities and losses didn’t have to happen. It was a waste of human life. A disaster that could have been avoided. The result of poor measures, poor trending, and reacting to random variation. The result? A real failure that could have been avoided if proper measures and trending had been in place.

So we are convinced … reacting to random variation has a negative impact. But what if you aren’t convinced? Then you still don’t understand the impact of natural variation. And you need to play the marble dropping game developed by Dr. Lloyd S. Nelson and described in Dr. W. Edward Deming’s book, Out of Crisis (MIT, 1982).

Would you like to learn more about trending? Attend the Advanced Trending Techniques Course just prior to the Summit.

Show Comments

Leave a Reply

Your email address will not be published. Required fields are marked *