David Schwinn ponders the dearth of understanding of statistics among managers, who have often had little training in statistical methods. Using statistics intelligently, understanding variation, and paying attention to the right numbers, he says, will assure better decisions and improved productivity. Hear him out…

As I titled this column, I was reminded that Dr. Deming liked to say, “The most important numbers are unknown and unknowable.” But some numbers are important, and most managers do not know how to manage them. I hope this month I don’t sound like a complainer, but this issue has been close to my heart for nearly thirty years and it arose powerfully for me as I got ready to teach a statistics course I haven’t taught in an academic environment before. I’ll share my story so you can get a sense of my biases and passions for this topic.

I don’t remember formally learning anything about statistics or variability during my undergraduate engineering education. We could find a formula for nearly every situation and get an exact answer. That ended with my senior design project when we were told to use all those formulas and get those exact answers, then multiply the answers by a “design factor” of 200-400%. That was for safety I thought, but, in retrospect, some of that had to do with the fact that up to that point we had unknowingly ignored the variation inherent in the design-and-build processes. I then had to complete my final thesis in which I generated quantitative prediction models about machine downtime. My advisor showed me how to calculate and include statistical confidence limits. I used them as I was told, without really understanding where they came from.

In graduate school, I took courses in business statistics and in operations research. Given that I could barely understand our instructor who spoke with a heavy foreign accent, I got through the course just fine by plugging and cranking the formulae, just as I had done in engineering school. I still did not understand the underlying theory. At Ford, when our corporate operations research folks could not solve a big operations problem, Dr. Deming solved it in ten minutes with a pencil, paper, and a control chart. Years later, when my wife, Carole, and I were working with Russ Ackoff, the co-author of the pioneering textbook on operations research, he declared that operations research in America was dead.

As I review my early career in engineering, management, quality, manufacturing, design, and research, I can remember only two areas where we used statistics. We unknowingly used inferential statistics to inspect incoming parts and we used control charts in another office to monitor and try to improve the quality of a different set of incoming parts. Dr. Deming and others later taught me that managers need to understand analytic studies in statistics. The primary tool for those studies, of course, is a control chart. I did not learn about control charts in school. Dr. Deming, Professor David Chambers, and a few others taught me the power of analytic studies to influence the future, versus the power of enumerative statistics (essentially descriptive and inferential statistics) to make a decision about what you have in your hand…great for deciding whether or not to accept a shipment of material.

At Ford, we started applying Dr. Deming’s theory on a grand scale. After I left Ford, I helped other manufacturing organizations use control charts to improve their operations. After a couple of years, we started to help manufacturing organizations to improve things beyond the factory floor, such as usage of resources, timeliness of orders and estimates, and employee turnover. We used control charts to reduce student dropout rates in educational institutions, reduce teller wait times in banks, reduce repair time in government military facilities, reduce resource usage at utilities, streamline benefits-providing processes at insurance companies, and increase sales at printers, stockbrokerages, and banks. All the while, we noticed plenty of evidence in the media that most people did not understand basic variation…two or three points does not a trend make. Finally, I’m ready to tell you about my most recent reminder of my concern.

I have been asked to teach Managerial Statistics for the first time this fall. The course has been offered for some years, but this is the first time I have been asked to teach it. I found both the syllabus and the textbook to be disappointing because of the heavy emphasis on enumerative studies. I also found the textbook to be very expensive. Remembering the series of “Statistics for Statisticians” workshops Dr. Deming used to lead toward the end of the 20^{th} century and the lines of academicians who emotionally confessed to Dr. Deming that they had never adequately explained the limitations of enumerative studies to their students, I expected to easily find an alternative textbook. As I contacted academic colleagues, I found that they have frustrations similar to mine with no easy solutions. Again, I was deeply disappointed. I was forced to create a course pack. Thank goodness, I had PQ Systems’ *Transformation of American Industry* and *Total Quality Transformation* materials to lean on. All this gets me to the point of this column.

I have now committed to teach managerial statistics in a way that, at least based on my own experience, truly helps my students–future and current managers–to use statistics intelligently to make decisions. I am asking you to do the same with the managers of your organizations. Every manager watches numbers. Please help them to do so intelligently and help them to figure out which numbers to pay attention to. I expect some of the managers in your organizations already do use analytic studies of data with which to make intelligent decisions. If you can share such examples with me, I would love to share them with my students. That’s it. Please help your own organization make better decisions and help me better teach my students how some managers already make such decisions.

I am particularly interested in your feedback to this column. It will be useful to try to achieve Dr. Deming’s purpose for the transformation of western management. You can reach me by commenting below.

I agree wholeheartedly with your premise of the “dearth of understanding of statistics among managers.” I took a business statistics course in college when I was 18 years old. With minimal work experience at the time, I had no frame of reference and it seemed irrelevant. Decades later, I now think it is the most underutilized “tool” in the manager’s toolbox. I’ve worked for many organizations—goverment, business, and nonprofit–and my impression is that none really use statistics to their full advantage, presumably because of a lack of understanding. My post-baccalaureate certificate and masters programs involved final projects using statistics but (thankfully to me at the time), my instructors drove my test selection and results interpretation. It is only now in my quest to upgrade from a Lean Six Sigma green belt to a black belt that I really understand the purpose and value of statistics. There is a seemingly endless list of concepts to grasp but it’s worth it!

Jane–Thanks for your response, reinforcing the idea that business education has not provided the statistical background, nor the understanding of the importance of data analysis. Anyone out there from a business college?

My experience in the UK is very similar – most management teams have no appreciation of variation and I never see control charts being used to track processes and metrics. There are lots of numbers available but no analysis, no understanding of whether processes are stable or unstable, no clue that many of the decisions they make are wrong. I teach project teams how to construct control charts and to use these to coach their managers on the power of the data that is in front of them. The data will always tell you a story, you just have to ask it the right questions.

As for a book on control charts, I would highly recommend Don Wheeler’s excellent “Understanding Variation”

You’re right, Mike—Wheeler’s the best, when it comes to understanding variation and its role in process improvement. Sorry to hear that things are no better in the UK.

More agreement here (unfortunately). In my experience, understanding of statistical concepts seems generally poor in organisations, resulting in low adoption rates of the tools. Quite revealing really when you are confronted with an issue which, after some basic investigation, is clearly the result of a lack of understanding of simple variation (and its causes). I’m guessing that the jargon and terminology, while seeming appropriate and correct to statisticians, contributes to the lack of understanding. A challenge for us all to tackle.

You’re right about the challenging terminology of statistics. People seem to be either scared to death or bored to tears at the prospect of studying statistics.

Thanks for the article David. I believe Six Sigma closes the gap that existed beforehand. The problem is, it is the Black Belts using the stats, not the managers. I believe GE had it right – progress the Black Belts to managerial positions, and change happens. Unfortunately the transition to statistically thinking management takes much longer than any of us had hoped.

As long as they can avoid data analysis, they can get away with it. Six Sigma and other initiatives help to put pressure on decision makers, and this pressure needs to be consistently applied.

Hi, I work in a small firm and it is really surprising to note that the trend is same everywhere in that people tend to avoid data analysis. I face the same issue wherein it is very difficult to get people start believing in data and focussing more on analysing them to find out the real cause rather going by general understanding of the data and using one’s intuition. Though I myself am no expert with statistics or the various sigma tools, but I do use data a lot to analyze the situation and try to use whatever I have learnt through my limited exposure and that really helps me to get a better picutre of the problem faster than any of my colleagues.

This points to the need for training at all levels, as Deming advises. Once people understand variation, they’re sold on data analysis, it seems.