figure 1: Control chart

In the 1930's Walter Shewhart remarked that the behaviour of physical phenomena and objects can be predicted more or less, depending on known causes of behaviour, while we have to use statistics to predict the behaviour which depends on unknown or unknowable causes. To predict the quality of products being produced he devised control charts with 3-sigma bands (figure 1). If some quality aspect of a product is within these bands, the variation is probably caused by unknown or unknowable causes (Deming called these "common causes"). If the quality aspect is outside these bands, the deviation is probably caused by something we can and should economically analyse and solve (Deming called this "special causes"). In order to get less 'natural' variation, one has to change the whole system that causes the variation.

An important attribute of quality is determined by the human appreciation of the product like for example: does the user of the product "like" it, find it "easy to use" or "easy to understand", or what makes a person decide to buy. Shewhart decided that many of these quality attributes are immeasurable in production and therefore only can be measured by trial and error. Trial and error doesn't have to be haphazard, but can be systematically researched. This led to what Deming at his speech to Japanese management in 1950 called the Shewhart cycle (figure 2):

figure 2: Shewhart cycle, as found in the transcript of Deming's speech of 1950 in Japan. Note that the person recording the speech didn't really understand it, witness the odd order of numbering of the phases.

Design-Manufacture-Sell-Observe-Redesign-ad infinitum. Continuous learning what makes the quality of the product better in the eyes and the feeling of the users, at the same time following the gradual change in appreciation of quality attributes in time. After all, what people appreciate tomorrow may be different from what people appreciate today.

Note that learning is only possible if we gradually and in small steps change the design in the consecutive cycles. If we change too much, some elements of change may be improvements and some may not, but we won't know which caused what. If we change in small steps, changing "one thing at the time" we can learn more about individual causes. If we then go through the cycle very quickly and frequently, we still can quickly adapt towards the most important and appreciated needs of the users, as well as the changing circumstances.

Deming, who was asked to present his speech in Japan in 1950 because Shewhart was ill, later made the cycle more general, wider than only for manufacturing, and called it Plan-Do-Study-Act (PDSA), while in Japan and at other places we also see Plan-Do-Check-Act (PDCA). Imai incorporated it into Kaizen, or continuous improvement, which of course is the same.

figure 3: Intuitive cycle

When I started using PDCA, I thought: Do is no problem, we do it all day. Plan we do more or less, usually less and for Check and Act we don't have time.
This is what I call the "Intuitive Cycle": Pl-Do-Pl-Do, as shown in figure 3. I cannot call it Plan, therefore I call it Pl: What is the next thing I should do, and I'm already doing it. This is how we live and work. We see something to do, and intuitively we already do it just as we did it before. If we would have to make a QFD (Quality Function Deployment) table every time we get hungry: "What are the possible solutions to solve my hunger, and which solution is the best Return on Investment", then we will be starved to death by the time the table is ready. In practice, we see or make some food, eat it, and the problem is solved.

If intuition would be perfect, everything we do would be perfect. Not everything we do is perfect, so apparently sometimes our intuition points us into the wrong direction.
So, instead of following the Pl-Do-cycle, let's first (see figure 4) Plan. Plan is twofold: What do we want to achieve ('the product') and how do we think we can achieve it the best way ('the project'). Then we Do according to the Plan, which is the first pitfall, because it means that the Plan must be Do-able and we must follow the Plan. If we assume this is the case, we can, after the Do, go to the Check or Study phase where we analyse: is what we achieved according to the Plan and is the way we achieved it according to the Plan? If yes: should and can we do it better the next time? If no: should and can we do it better the next time? Then, in the Act phase we are going to decide what we will do differently the next time, because if we don't do anything differently the result will be the same.

figure 4: Plan-Do-Check-Act cycle

And, as said, we do very small steps at the time, so that we can learn what is clearly better and what is not. Doing these small steps very quickly and frequently, we still can improve quickly what we are doing (the product), how we are doing it (the project) and even how we organize all that (the process).

Because in the Act phase we introduce mutations of what we do and how we do it, we call this the Evolutionary approach.

Gradually, I found that people often are 'stuck in the Check phase': People often quite well know what's not going so well, what's a bottleneck, and easily say: Yes, we should do that, but ... (management doesn't allow it, it can't be done, etc). If I hear these "Yes, but ..."s I always say "You're stuck in the Check phase. If that really is a problem, what should and could we do about it?" and the other person always immediately has a suggestion what we should and could do about it. The problems we face are not the real problem. The main problem is that we have to move to do something about it. Not staying in the Check phase, but rather moving to the Act phase: What are we going to do about it. Not complaining what cannot be done, but rather think what can be done. Then you'll see that much more is possible than we thought.