Home ] Up ]

Excerpts from: Ray Stata, Organizational Learning ¾ The Key to Management Innovation, Sloan Management Review, Volume 30, Number 3, Spring 1989, pp 63-74:


Quality Improvement:

A Methodology for Change

Even when there is a strong consensus for change, achieving it is easier said than done. For example, another imperative reinforced by the planning process at Analog Devices was the need to improve customer service, product quality, and yields. Of course this concern was not new. Since the early 1980s, as our customers have gotten a taste of what Japanese electronic companies could deliver, and as just-in-time JIT) programs have become more prevalent, pressure has mounted to improve performance. What was new was the realization of just how much we had to 'improve to meet our customers' expectations, and how little time we had to do it. On-time delivery of products that work has become the major factor in vendor selection and performance evaluation. We can no longer win by the sheer force of being first to market with the latest products and technology.

Quality improvement, or total quality control as it is often called, is a management methodology for achieving improvement and change. In 1983 we began to introduce quality improvement methods at Analog Devices. We decided to focus our attention on product quality, on-time delivery, lead-time, yields, and new-product time to market. We went to seminars, read books, gave speeches, and introduced information systems to measure our performance. But three years into the mission we were not getting very far very fast. I had an uneasy feeling that I did not know what I was supposed to be doing to lead this effort and that there were a lot of other dedicated managers in the same boat.

We knew all about error detection and correction and about doing it right the first time. But we did not have any notion of what rate of improvement was satisfactory or what we could do to accelerate the 'improvement process. Considering that many Japanese companies had been working the quality improvement game for more than twenty years and that they are not standing still even now, we had a Justifiable sense of discomfort.

Because of our "lean and mean" attitude toward staff functions, we had resisted the addition of a quality improvement staff. Line managers were expected to learn on their own. But learn from where, learn from whom? Reading books and going to seminars was not enough. So we finally broke down and recruited a quality improvement professional to teach us how to tap the mainstream of experience and knowledge that is rapidly accumulating in this field and to help our managers become more expert practitioners. Only then did the organization begin to see real progress. One of the early lessons I learned from our quality guru was that there is a rational basis on which to set standards for rates of improvement. From his consulting experience, our director of quality improvement had documented case histories where quality improvement methodology had worked. What these cases showed was that while the rate of improvement varies from case to case, the rate in each case is remarkably consistent over an extended time period. Figure 2 shows three actual businesses' learning rates. Note that in the first case, performance improved by 50 percent every 10.4 months, 'in the second case every 7.8 months, and in the third case every 3.6 months. He called this characteristic slope of improvement the half life.


An analysis of a larger number of case studies indicated that the half life for improvement fell within a relatively narrow range, usually six to twelve months, across a wide range of applications.' The reason for this phenomenon is clear enough when you understand the method by which quality improvement is achieved.

The method is deceptively simple. For example, as I mentioned, one of our goals was to reduce the percentage of orders shipped late. To do this we assembled a team from various organizations involved with customer service to analyze the causes of lateness. For each late shipment we determined the cause, and then we plotted their distribution. We found that a relatively small number of causes was responsible for 50 percent of the problems.

Next we assembled problem-solving teams to attack these major causes of lateness. When the cycle was completed, we repeated the process by prioritizing the causes for So percent of the remaining problems and then eliminating those causes. This cycle was repeated again and again; each time the most important remaining problems were identified and resources were focused on solving them.

In this example, as in others using this method, the slope of the learning curve is determined by how long it takes to identify and prioritize the causes of the problem and to eliminate those causes. The skills of the people and the level of resources do have an impact, but surprisingly the time required for each cycle of improvement is largely a function of the complexity and bureaucracy of the organization. Or, to put quality improvement in the larger context of this paper, the slope of the characteristic half-life curve is determined by the rate of organizational learning.

Notice that this theory of learning differs from the Boston Consulting Group (BCG) "experience curve" theory that says learning occurs as a function of cumulative production volume, independent of lapsed time. The quality improvement theory says that learning, properly managed, occurs as a function of time, independent of cumulative volume. How else can we explain the success of the Japanese automobile industry, which learned faster than the U.S. industry with substantially less cumulative volume? If we combine the two ideas, we can say more accurately that the slope of the BCG experience curve is determined by the rate of organizational learning. A steeper experience curve occurring at lower production volume can soon overcome a more shallow experience curve occurring at higher volume.

We know that communication across organizational boundaries is less effective than within organizational boundaries and that many problems accumulate because of poor communication. Quality improvement is a way to create temporary organizational structures, or teams, that cut horizontally across organizational boundaries and enhance communication and cooperation. It is a way to get people to think about problems and issues objectively and quantitatively instead of subjectively and politically. It is a way to separate the vital few problems from the trivial many-and to focus organizational resources on resolving them. In short, quality improvement is a way to accelerate organizational learning.

Using the half-life concept, at Analog Devices we set very aggressive five-year goals for quality Improvement (see Table 1). The results of continuous improvement with nine-to-twelve-month half lives over an extended period are awesome. The first reaction of our organization was to recoil from what looked like unrealistic objectives. But we reminded our managers that if a company really gets its quality improvement act together, there 'is no fundamental reason why these goals cannot be achieved. There are companies in Japan already operating at these levels on some of these measures. …

…Another problem with management information systems is that they are strongly biased toward reporting financial information to stockholders and government agencies. Unless quality improvement and other more fundamental performance measures are elevated to the same level of importance as financial measures, when conflicts arise, financial considerations win out. To address this issue, we designed what we call a division scorecard that reports only the barest of financial information and places greater emphasis on quality improvement goals. This scorecard is used not only to evaluate division performance, but also to structure division bonus plans.

How information is displayed makes an incredible difference. Consider, for example, the format we use to display on-time delivery information (see Figure 3). This simple summary replaces pages of information that used to be circulated to managers. With all these pages, the most crucial information was missing-namely, the half-life trend. Because the information is plotted on a log-linear scale, the trend is readily discernible. For management purposes, displaying all divisions together on a single page has great motivational value. A high level of internal competition exists to generate the fastest learning curve; it is obvious and embarrassing when you are not performing.

Management information systems transform data into information and then help managers transform information into knowledge and knowledge into action. The challenge is deciding what information and knowledge-in what form-are needed. If we keep organizational learning in mind as a goal of information systems design, then we are more likely to generate the information and knowledge that managers need to take effective action.”




back to top

©1999-2006, Arthur M. Schneiderman  All Rights Reserved

Last modified: August 13, 2006