On Facts vs. Opinions

In lean, as well as in many other areas, facts hold much more value than opinions. But what are facts, and what are opinions? Unfortunately, this is not black and white, but a big gray area. This blog post is based on a question by a reader on my earlier post “Why It Is So Important for Managers to Listen!” Let’s dive deeper into how to distinguish facts from fiction, and how to get facts in the first place…

The Radical Approach: Cogito Ergo Sum

Let’s start with the ancient Chinese text “Zhuangzi” by Zhuang Zhou, where he dreamt about being a butterfly. After waking up, he wondered if he was indeed a human dreaming about being a butterfly, or if in reality he was a butterfly dreaming about being a human.

The French philosopher René Descartes (1596–1650) used a similar pure philosophical approach, even though this is completely impractical for reality. He developed the first principle “cogito, ergo sum,” usually translated into English as “I think, therefore I am.” In this approach, he decided to doubt everything, including his eyes, ears, and other senses. As a result, he concluded only that he must exist in some form, because he is thinking. Everything else is uncertain.

Other philosophers even expanded on that idea even more radically, questioning if it proves that there is an “I,” and concluded that we can only say for sure that “something is thinking,” and there is no certainty whether there is really an “I.”

Again, this is completely impractical for reality, and I encourage you to (usually) trust your eyes and ears. Albeit, you still should be careful, and I leave it up to you if this portrait of Descartes with his very strong legs is accurate, or if an AI merely expanded the well known portrait of Descartes and decided that there should be legs…

The Scientific Approach: Observe/Measure and Replicate

Different from philosophy, science tries to understand the world. As such, scientists base their conclusions on observations and especially measurements. A scientists measures his or her observations, and based on these observations makes conclusions about the underlying reasons behind.

The key point here is that other scientists must be able to replicate these findings. The scientific literature is full of papers claiming results, only to have other scientists be unable to reproduce these results because the science of the first paper was flawed. (Errors, methodical flaws, lack of statistical significance, or even intentional data manipulation are common. One well known example is cold fusion. Another is the cure of cancer, which pops up frequently in the news.)

A scientific finding is accepted if it can be consistently replicated and is statistically relevant. This also means that the more fluctuations you have, the harder it is to create a valid finding. Because of this, findings in social sciences are often much more fudge-y than in other sciences that can measure with higher precision and repeatability.

The Practical Approach: Observe/Measure

On the shop floor, you can also measure a lot of things, albeit also often with significant fluctuations. The goal, however, is not to find the ultimate truth, but to improve your system to make (more) money. Hence, relocations are often too expensive even if they would be possible. It is rare that a colleague in the company repeats your measurements to confirm that they are correct (and if a repetition happens, then it’s usually to blame someone and attack their ideas…).

Hence, when talking about facts in lean manufacturing, we usually mean measurements, observations, and statistics. However, unlike in science, to save money many measurements are not repeated to reduce cost and to increase the speed of implementation.

The Better Practical Approach: Observe/Measure, and Repeat if Results are Wrong

Since there is a lot of pressure to implement improvements quickly, sometimes mistakes do happen. But even in the interest of speed and cost, do not skimp on a proper analysis. The Toyota Practical Problem Solving (PPS) puts most effort into its analysis. Or, in lean terms, the “Plan” of the “Plan-Do-Check-Act” PDCA should be done thoroughly.

The last step of the PDCA, the “Act,” is also a sort-of fail-safe to catch mistakes. If the implementation does not yield the desired results, THEN you have to measure and analyze again to see where you went wrong. In other words, in manufacturing, you repeat the measurements or analysis only if the results are unsatisfactory.

A Common Approach: Let Someone Else Observe/Measure

Observing and measuring takes time. Sometimes it may be easier to let someone else make the observations or measurements. This could be another employee, sometimes even the operator, or even a computer system. This could save a lot of time, or at least use the time of another (lower paid?) employee.

On the other hand, this introduces the possibility of inaccuracies. Similar to you doing the measurements yourself, there may be misunderstandings or wrong assumptions and simplifications. However, with someone else doing the observations, measurements, and the subsequent analysis, there may be additional issues.

First, for many analyses, you need to make assumptions and simplifications. If you do this yourself, you probably understand these assumptions and simplifications. If someone else does it for you, you may not know the underlying assumptions, and you may make the wrong conclusion.

Second, often, only the averages are presented. If one number is bigger than the other, the conclusion looks obvious. However, in most cases there is considerable uncertainty about the exact values due to the fluctuations in measurements. In science, usually a 95% confidence is needed to make sure that one value is indeed larger than another one, giving many results that are not statistically significant. In manufacturing, a 95% confidence may be too harsh. In fact, these statistics are usually not even calculated. But by looking at the raw data, you can get a feeling if two values are truly different, or if it may be just random flukes. Suddenly, a clear answer for one option may turn into an “either way works” result.

Finally, there is intentional manipulation. This happens both in science and also in industry, where one person wants to influence the outcome in his favor. This occurs way more often than you would believe, with all kinds of tricks. Cherry picking the data, massaging the equations, showing only selected “truths” while omitting others, or even publishing outright lies—there are many ways to manipulate the numbers. For the OEE, I even wrote a sarcastic article on the Top Three Methods on how to Fudge Your OEE.

Go to Gemba!

Overall, distinguishing facts from fiction is difficult, and there is a lot of gray area. You probably cannot measure everything yourself, and hence you have to trust data from others. In any case, you should at least sometimes verify whether you can observe this on the shop floor too, as described in my post Visit the Shop Floor or Your People Will Fool You! – Genchi Genbutsu. Hence, everybody related to manufacturing needs to go to the shop floor at least sometimes, i.e. go to Gemba! Now, go out, be a beautiful butterfly that dreams about lean manufacturing, and organize your industry!