Home > Gemba Coach> Is Lean Scientific?

Is Lean Scientific?

Permalink   |   3 Comments   |   Post a Comment   |  
  |   RSS

Dear Gemba Coach,

I often hear that lean is like the scientific method. As a lean consultant, I facilitate kaizen events, and find it hard to see the parallels. Could you clarify what people mean when they say this, and how it applies to our work?

Thank you for this very fundamental question. It is clear that the founders of lean thought they were applying the scientific method (in the form of Deming's PDCA cycle) to business problems. As Art Smalley and Durward Sobek point out in Understanding A3 Thinking, their great book on A3s, on page three of the first leaflet ever produced by Toyota on TPS, it says: "TPS is founded upon a scientific mindset. It is important to start with those phenomena and search for the true cause by tracing things to their origin. In other words, we place heavy importance upon getting the facts." Indeed, many misguided Gemba implementations can be linked to a failure to realize kaizen in the spirit of the scientific method. Still, the term "scientific method" is often vague and open to interpretation (for instance, Taylorism is also known as "scientific management," which is both true and untrue, but confusing). So, what are we talking about?

There are many aspects of the scientific method, but I would summarize (and over-simplify) it as follows:

  1. Observation: the first step of any scientific investigation is observing a phenomena in reality (as opposed to philosophy, which is essentially about mental arguments), which is to say in context.

  2. Formulating a theory: putting down on paper a clear cause-effect link that explains whatever you are observing. This is never easy because many phenomena seem complex. There are many causes and effects happening at once, so there is an art to reducing it to a simplified "this happens because of this . . ." Unless this phenomenon is well known previously, a theory is often nothing more than an educated guess on why things work out this way.

  3. Look for anomalies: this is where scientific thinking goes against common sense. Finding cases where our pet theories are shown to be correct is the easy part. True scientific thinking, however, is about finding conditions in which the theory doesn't work. Anomalies flag areas we can learn from, because they reveal where our model of reality is not good enough to explain what really happens.

  4. Defining the problem: expressing specifically how anomalies diverge from the stated theory is the real lynchpin of scientific thinking - this is where learning starts.

  5. Formulating hypotheses: once the problem has been clearly stated, several hypotheses can be explored to explain this problem. The idea at this stage is not to find a quick answer, one which makes the problem go away, but to find the right answer, which reconciles (or not) the anomaly with our general theory.

  6. Experimenting to test hypotheses: repetitive testing of the various hypotheses in live conditions is what leads to solving the problem.

  7. Reformulating the general theory to incorporate the newly validated hypothesis, which is, by and large how our learning grows.

  8. Sharing this knowledge, not just by explaining the results, but by allowing others to replicate the experiment in different conditions in order to replicate the learning and further the understanding of the new theory.

This sounds terribly cumbersome (and is), but since the late eighteenth century, this process, and the interaction between scientists and engineers has been the main source of industrial productivity. And this is the approach that Toyota production engineers adopted when they heard it from Professor Deming. In lean terms, this is how it is applied:

  1. Go and see: genchi genbutsu is the foundational practice of lean management - going to see the facts for yourself at the source. When problem solving fails, more often than not, this is a failure of "go and see." We have not spent enough time at the real place watching the process unfold to figure out what is really going on. Just recently, I was discussing with a problem-solving team how to solve a complex quality problem, and as we discussed the problem in the production cell, we saw the operators being interrupted almost every five parts by something or other. A third of these interruptions occurred at the workstation where the quality problem occurred. We had been trying to figure out where the machine cycle went wrong, but then we focused on looking at how the operator started working again after just being interrupted in her cycle. Just by observing in this way, we could see several quality risks happening as she picked up the work and finished the cycle - none of which had anything to do with the machine's process.

  2. Write the standard: writing a standard in lean terminology is the equivalent to formulating a theory. Writing the standard means putting down on paper the theoretical sequence of operations necessary to do the work right. This is not so easy. In many cases, we have to look for the original documentation, and we immediately see that many parameters have drifted from what was specified in start-up conditions. The first step here to solve the problem is to bring the process back to initial conditions (fit to the theory) and see what happens.

  3. Look for variation: any variation from standard is an anomaly in the process. Once the standard is clear, rather than tell ourselves that "normally" the process is this or that, we focus on every variation that we spot, anomalies in scientific terms, and try to understand them.

  4. Defining the problem: a problem, in lean, is a gap with a standard. This gap is expressed in two ways: first, as a performance gap with a standard (such as the best day, or the best plant) and second as a variation from the standard process (what we need to do for the work to be right). In practice, there are often all kinds of variation, and the skill in defining the problem often lies in identifying which variation has the greatest impact on the performance gap. This is not done by thinking hard, but by testing every factor of variation until we're quite sure which of them has the biggest impact.

  5. Seeking the root cause: once we've figured out (through trials) the most important source of variation, we must determine the root cause of this variation. This is where the '5 why?' method is so powerful. It enables us to move away from superficial explanations and delve deeper in our understanding of the process.

  6. kaizen: kaizen is often nothing more than "try it!" The purpose for kaizen is to try various things to figure out the real cause of the problem - and so be able to fix it once and for all. Although the scientific process is cumbersome, the experimentation par of it needs to be quick, plentiful, and tireless. In his attempts to create a light bulb, Edison is supposed to have quipped: "I have not failed. I've just found 1,000 ways that won't work."

  7. Reformulate the standard: Once the problem has been solved, because the root cause has been identified and the proper countermeasure implemented, the new standard is formulated and put in place. As with scientific experimentation, a new standard often also means changing the organization around the activity.

  8. Yokoten: in this, lean practice is remarkably similar to the scientific method. Rather than spreading the "best practice," Yokoten is more about replicating the study in other processes rather than applying the solution. One need not start from scratch, but by reproducing the same experiment (kaizen) in different conditions, one is likely to end up with different applications – and hence better local learning.

One of the more frustrating aspects of working with old style sensei is that they never seem to be interested in the solution - the improvement actions. All their questions are about understanding the problem. At first, I thought that this was a kind of trick to impress the troops. But as time went by I came to accept that they were genuinely not interested in any solution per se: they trusted people to come up with what they thought was best. They really were interested in understanding the root cause better, in learning by deepening the collective understanding of the situation.

Shades of Gray

Realizing how much lean draws from the scientific method has profound Gemba implications. Lean practice is mostly about finding and addressing gaps between our current processes and our ideal ones; and not about spreading solutions. One typical example is stock. Management wants to reduce inventory. Lean thinkers want to reduce lead-time. Reducing inventory can be achieved by decree (with sometimes dire results in terms of on-time delivery). Reducing lead-time can't be forced: the causes of lead-time (large batches, planning issues, complex flows, poor logistics, etc.) must be understood and tackled one by one on specifics. If this misunderstanding is not cleared up, many people look for answers that aren't there. Lean "solutions" tend to be no more than generic ways of solving generic problems (use a small train to deliver parts to the line), but specific solutions (here are the wagons you should use) are rarely workable in real life circumstances. Trying to apply literally solutions from other sites or companies often leads to bitter disappointments (as well as unnecessary expenditures).

Lean answers are seldom black and white. When people catch religion from reading the lean literature, they try to apply across the board lean "practices." Because it follows the scientific method, by their very nature lean countermeasures (no one ever believes in full blown "solutions") tend to be context dependent and, essentially, shades of gray. For instance, in the case of bought-out-parts inventory, people who've convinced themselves of the superiority of "flat storage" often want to pull all the racks down. It's sometimes hard to convince them that pulling racks down to store components in supermarket-like shelves is a great idea, but some parts do have long lead-times if they come halfway across the globe, and some other parts are dead stock but not chalked up for destruction yet and so on. So a few racks are necessary, and these racks will probably be managed with a traditional computer ordained address system, contrasting with the supermarket-type handling of the flat storage. It's true that the full gain will only appear when all forklifts can be pulled out of the hall, but this requires solving the problem of long lead-time parts and slow moving parts. Different principles apply in different conditions.

kaizen Rhythm

There are three broad practical consequences from trying to apply lean in the spirit of the scientific method. First, we need to spell out our theories about how things work continuously, whether as standards or causal relationships. We must see the cause-and-effect relationships within processes. As the first TPS leaflet states: "in problem solving, the purpose must be made clear ... in kaizen the needs must be made clear." This is not always obvious, but it's a worthwhile exercise, and key to progress over time. Secondly, we need to increase the rhythm of kaizen activities and events. If learning is about failing, we must fail more, and quicker. People must learn how to try things out right away in a safe manner – with cardboard cutouts, duct tape, whatever. Quick experiments yield far much more knowledge than hours of discussion. Finally, we must abandon our belief in cookie-cutter solutions and accept that if we want another group of people to apply a specific improvement it's both safer and quicker to get them to replicate the analysis and kaizen process rather than to impose our specific solution for implementation. All knowledge is contextual, and no conditions are ever exactly the same.

kaizen, kaizen, kaizen. I can't stress enough the need to maintain a high rhythm of kaizen activity. The industrial revolution became a revolution (as opposed to slow evolution since the thirteenth century) when the rate of theory formulation and experimentation skyrocketed. The same is true of lean progress within one company. Lean transformation becomes a transformation (as opposed to business as usual with a little continuous improvement thrown in) when the rate of kaizen and problem solving explodes. At the very least, every supervisor conducts a kaizen event a month, and every middle-manager works on one A3 always. Lean is work! To quote Edison again: "Opportunity is missed by most people because it is dressed in overalls and looks like work." So is improvement.

3 Comments | Post a Comment
Jamie Flinchbaugh December 29, 2009
I think many great lean books draw significant linkages between PDCA and lean (assuming you understand PDCA to be linked to the scientific method) such as Durward and Art's book, as well as Tom Jackson's hoshin kanri book and much of Steven Spear's writing where he describes Toyota as a community of scientists.

I think that lean thinkers should more directly import the language and practices of the scientific method into lean work. Too often kaizen is done based on assumptions and untested 'improvement' ideas. I strongly advocated building theories or models and taking the time to develop a hypothesis. I believe this is what builds effective learning organizations.

Jamie Flinchbaugh
Olivier Fichet January 11, 2010
M. Ballé has described how the "problem solving" dimension of LEAN, when rolled out with rigour (per Jamie's comment on PDCA)is scientific by nature. Indeed, conducting improvement of material/physical processes based on objective principles involving the systematized observation of and experiment with phenomena is Science.
I think there is a further reason why LEAN is scientific. It has to deal with the other main dimension of LEAN: "Just In Time". With Just In Time approach -I almost would like to write: JIT theory-, one can systematically implement a process that optimizes the needed resources to achieve a targetted output. The math is known and well described. It is based on actual demand, desired service level and observed cycle times. And as well, current performance in Quality, and supplier's Lead Time. The results can be calculated or elaborated and are: standard work (manpower and tooling), inventory levels (material)and capacity needed (number of machines). In traditional environments (non LEAN), the same manpower, material levels and needed capacity are sometimes obtained with far less structured approaches, sometimes not even integrated in the same framework. For example, there is a common "rule" in ERP run operations that one should have no more of one week of inventory for the few highest cost components representing 80% of the components value, two weeks for the next 10% in value and 4 weeks for the last 10% in value...this is financially driven, and can prove a recipe for failure at operational level. Or one may "staff" a production cell including a "manpower efficiency" factor, which once again has little to do with the production modell. Or one may base inventories on pure marketing forecasts rather than history of actual sales...
SMED or TPM are as well scientific tools. These are approaches or patterns that will systematically drive to improvements in change-over, resp. machine efficiency/OEE.
In conclusion of this comment, LEAN is scientific as well because it is (more and more thanks to LEI Authors) a systematic and formulated knowledge for "wasteless transformation and delivery of goods and services".
John Foster February 15, 2012
What is Quick Response Quality Control and how does it differ from other problem solving methods?
Other Michael Ballé Related Content

Gold Mine Master Class



  • What is your psychology of change?
    Dear Gemba Coach,  
    Do you have a psychology of change? And if so, what is it? Where should we start?
  • Can I change a company's attitude that people can't be trusted to do their jobs unless they are scared into complying?
    Dear Gemba Coach,  
    A major assumption in lean thinking is (unless I’ve got it all wrong) is that people genuinely want to do a good job, and the only thing standing in their way is a poor system. In other words, it’s the assumption that most people have high inner motivation. But some (well, probably many) organizations act on the assumption that you can’t really trust anyone to do their job unless they are constantly controlled and scared into complying. It’s hard for me to see how lean can help any org without changing this assumption first ... or?
  • Why don't I see any significant performance improvement from obeya rooms?
    Dear Gemba Coach,  
    We’ve deployed obeyas all across our organization, but I can’t see any significant improvement in our results. We can better see which teams perform and which don’t, but the good teams stay good and the poor ones poor and I’m not sure the increased performance is worth the effort – are we doing something wrong?