Jan 9 2013
Last week, my Suggested Content on Scoop.It! contained a link to a May, 2010 working paper from MIT’s Engineering Systems Division (ESD) by Kirkor Bozdogan, entitled Towards an integration of the Lean enterprise system, total quality management, six sigma, and related enterprise process improvement methods . For a scoop, it is a bit stale but it nonetheless caught my attention and I would like to supplement Bozdogan’s academic perspective with my implementation experience.
Describing these approaches as “complementary,” as Bozdogan does, avoids controversy, but I don’t believe it is accurate. They really are competitive brands put out by consultants vying for clients in overlapping markets. And they are so different is scope and track record that they do not belong together in a list. The topical literature is a cacophony of claims of effectiveness, originality, and universality, as well as bandwagon jumping. It befits a marketplace, and competitors should be expected to pitch aggressively. As Hillary Clinton said about running for office, “you cannot be above the fray, it is a fray.” It behooves the clients to sort the wheat from the chaff and make their choices.
Exposure to the Toyota Production System (TPS) sparked my interest in manufacturing, in Japan in 1980, but then I immediately went to work in the semiconductor industry, where TPS is not much of a fit. Over the years, I have been exposed to all of the approaches surveyed in the article, and formed opinions about them, that I am sharing here. In particular, I would like to explain why I chose to work under the Lean flag and none of the others.
The sequence of topics is as follows:
- Brands versus Science
- Market share
- Different approaches for different enterprises
- Change programs in corporations
- The “Methods” and their descriptions
- Conclusions on the consulting profession
As academics are prone to do, Bozdogan treats Lean, TQM, Six Sigma, etc., as if they were scientific theories, when in fact they are marketing brands, developed by consultants for commercial purposes. It is not quantum versus Newtonian mechanics, but Coke versus Pepsi. An engineer or a manager conducts a successful project within a company, and, in doing so, develops a unique approach to a problem. The next step may be to leave the company, go into consulting, and market this approach as a more general solution, but it needs a name to distinguish it from competing offerings. If successful, it is then diluted for application to ever broader classes of problems. Five years on, PhD students are writing dissertations about its theoretical foundations; ten years on, articles appear in business magazines bemoaning its failure to deliver the expected results.
Sometimes, the name is coined before the content is developed. A well-known marketing story is that the name Hershey’s Hugs was an entry by a Hershey employee in a naming contest. The product, a Hershey’s Kiss with a twist of white chocolate and milk chocolate, was then developed to match the name. This pattern was followed in manufacturing consulting in at least two cases. When Roger Nagel at Lehigh University created Agile Manufacturing in the early1990s, he defined it as the “next step after Lean,” whatever that might be. Likewise, I have heard a rumor that TPM was a label concocted at JMA in the late 1960s by a group of consultants who wanted to offer services in the Maintenance area, and that the content was developed afterwards.
This is not a criticism of consultants. They have to package their ideas in a way that attracts clients and inspires confidence, or else these ideas remain unheard and unimplemented. And it may take several attempts to get it right. The Toyota Production System was marketed in the 1980s as Just-In-Time, Demand-based and World-Class Manufacturing before John Krafcik came up with the Lean label. For Six Sigma, the title of “Black Belt” for engineers given basic training in statistical design of experiments was marketing genius, as it had none of the wussiness of “Staff Statistician” and implied a non-existent link to Japanese martial arts.
How successful are these brands in the market? One easy way to answer this question is to check LinkedIn discussion groups. LinkedIn is the largest professional social network, with >135M members worldwide, and the amount of chatter each approach generates among them strikes me as a valid measure of its influence as of 12/31/2012. I found the following:
- If you search LinkedIn groups for Lean, you find 2642 groups with the largest ones having tens of thousands of members. The memberships overlap, and therefore their counts do not add up, but largest group, Lean Six Sigma, has 160,000 members and, if you check the content of the discussions, you find that it all about Lean, not Six Sigma.
- Searching for TQM, you find 87 groups, with the largest having 942 members.
- For Six Sigma, you find 1411 groups, but nearly all are about Lean Six Sigma, which, contrary to what the name would lead you to believe, has next to no strictly Six Sigma content.
- The “related” methods also reviewed in the article are the Theory of constraints (TOC), Agile manufacturing, Business process reengineering (BPR).
- For TOC, you find 177 groups, topping out at 1307 members.
- For Agile manufacturing, 10 groups topping out at 2,700 members, but, for just Agile, you find 1481 groups with the largest having 37,572 members, reflecting that, even though the Agile brand is dead in manufacturing, it lives on in software development and project management.
- Business Process Reengineering (BPR) has 16 groups, topping out at 4,500 members, but these groups are not dedicated to this topic. They are about business processes, with reengineering listed among the topics of interest.
These numbers make a statement about the relative popularity of brands, not the technical effectiveness of approaches. It is possible for the best products to fail in a market for all sorts of reasons. But these are ideas and, in the market of ideas on management and technology, time has a way of filtering the bad ones. The newest of the approaches discussed in the MIT article, BPR, has been around since the early 1990s; all others, since the 1980s. If they have not made their mark by now, when will they? The inescapable conclusion is that Lean is the only one that is thriving. As a consequence, the bulk of the article should be about Lean and the reasons it crushed its competition.
The article treats “the enterprise” as a generic entity. The introduction refers to different types of enterprises, but the body of the article does not. Even within just Manufacturing, companies that make products through machining, fabrication and assembly have been under the influence of car makers for 100 years, first Ford, then GM, and now Toyota, as a result of which their management responds to the Lean message. By contrast, companies that primarily run a chemical process and package its output, like detergents makers, are much more focused on the maintenance of their facilities. They have been more receptive to TPM, and have then lumped under the TPM label all sorts of improvement activities that are not related to maintenance. In high-technology fabrication, as in the semiconductor industry, if your processes are mature, your products are obsolete, and you must therefore constantly face the challenge of producing in high volume with immature processes. This makes you receptive to the promise of Six Sigma and its techniques for enhancing process capability. “Services” covers an even broader range of activities, from engineering development to car rental, with different needs, both actual and perceived.
When the top management of a corporation embarks on a change program, it does not only impact the content of the work, but its organization and power structure as well. There are evolutionary paths to promotion in a corporation. Excel in sales and you rise in the sales hierarchy; get involved with a successful new product and it will make your career; exceed your goals as a production supervisor and you will on your way to become plant manager… This is the normal operation of the organization; it does not alter its structure. By contrast, successful change programs are revolutionary, in that not only are their supporters rapidly promoted but their opponents are brutally pushed aside, sometimes demoted, often fired. Change programs can degenerate in many ways, for example by attracting zealots who turn its original clever methods into a dogma that is used to exclude not only non-believers but heretics as well. This manifests itself in a variety of ways, from mandating the use of VSMs and Kaizen events in Lean, to making a Six Sigma black belt a requirement for promotion.
The politics make it difficult to even have a rational discussion on the merits, or even the relevance of a tool in a particular circumstance. And the presence of camps with interests in the success or failure of the program prevents the collection of objective metrics.
Bozdogan labels all the approaches as “enterprise process improvement methods,” which, at least in the case of Lean, strikes me as overly restrictive. First, it is not a method, in the sense that it does not have a sequence of 12 steps you can follow without thinking and expect to succeed. Instead, the implementation of Lean, particularly outside of car making, requires you to abstract the principles behind the Toyota tools and select, adapt or develop new tools to apply these principles in a different context. Second, the term process improvement implies an exclusive focus on how things are done, or tactics, as opposed to what things are done, or strategy. And Lean is a business strategy, requiring leadership and participation from top management, not just a tactical tool. It isn’t just about fixing details of operations, but also about make-versus-buy decisions, new plant and new line designs, human resources and compensation policies, etc.
The description of the Lean enterprise system in the paper (section 2.1.1.) is actually accurate, if too brief. While it is broader is scope and depth than any of the other “methods” covered, it received the shortest explanation. As explained in my author’s page, my involvement with it dates back to 1980. Lean is based on the Toyota Production System, and therefore has the following, unique characteristics:
- At least in Manufacturing, it encompasses all facets of the business. Toyota designs, develops, makes, and markets products and therefore has an approach to all these activities, as well as management and support functions like Maintenance, Quality Assurance, Accounting, or Human Resources.
- As Takahiro Fujimoto puts it, it is system that emerged over decades, as Toyota engineers and managers developed solutions to overcome crises as the company grew from nothing to the largest car company in the world. There is no theory of which it is an application. Instead, it is a living and evolving system, from which we have to reverse-engineer underlying principles in order to deploy them in other contexts.
Of course, we shouldn’t be blinded by enthusiasm. While learning from Toyota, we should not assume that it is perfect. We should not forget that it is a business, run by fallible human beings, and with a commercial interest in its image. We should not assume that its system applies everywhere to everything, and we should keep our eyes and minds open to equally good ideas from other outstanding companies, like Ikea, Apple, or Michelin, or from management and technology thinkers. And we should never blindly apply recipes but instead use what we learned to work through the specific issues of every factory.
Today, as an object of consulting, TQM is dead and little more than a historical footnote. Bozdogan’s description of TQM make no reference to TQC, and in particular the Japanese version of TQC, on which it is based. By 1975, all the major manufacturing companies in Japan had received the Deming prize for implementing TQC, and all that could be gained from it was incorporated in their practices. It was a successful approach that had run its course. TQM, as a watered-down version of the Japanese TQC, became the object of the Malcolm Baldridge National Quality Award in the US in 1988, and soon lost credibility as a result of being given to organizations that were notorious for bad quality. The spirit of TQM, however, lives on in the ISO-900x series of standards, for which certification has been a cost of doing business for many companies.
In his recollections of implementing Six Sigma at Allied Signal, Six Sigma creator Mikel Harry proudly recalls getting an executive fired for speaking up in a meeting in favor of the previous improvement program, Total Quality Leadership (TQL), the company’s version of TQM. So much for these programs being complementary!
If you peel away all that has been piled onto the original Six Sigma from Motorola in the 1980s in order to make it a universal approach to the enhancement of enterprise performance, what you find is a modernization of the SPC of the 1930s for the purpose of addressing the process capability issues of high-technology manufacturing. In the semiconductor industry, you first develop a process to make chips, and then design products to make by this process, which is not the way most other industries work. Then competitive pressure forces you to start producing chips in high volume at yields as low as 10%, and start a battle for yield enhancement that is key to market share and profitability. This battle ends two years later, to be rejoined immediately on the next generation of technology.
In the old SPC, a state of statistical control for a process variable of mean μ and standard deviation σ was arbitrarily defined as having its tolerance interval contain the [μ-3σ, μ+3σ]. It is was a normal variable, then 99.7% of its values would be within the tolerance interval. The problem that became critical with electronics in the 1980s was that, if a product had 100 such independent process variables, they would all be within their tolerance intervals .997100 = 74% of the time, meaning that the product would be 26% defective. Raise the requirement for each variable from ±3σ to ±6σ, and the ratio of values out of the tolerance interval drops from .3% to 3.4 ppm. Then, with 100 independent, normal process variables as before, the ratio of defect-free products goes to (1 – 3.4╳10-6 )100= 99.97%.
For non-normal variable, or for attributes, the meaning of Six Sigma extends to the achievement of <3.4 defects per million opportunities (dpmo). As we have seen, with as few as 100 defect opportunities in a product, it works out to .03% = 300ppm of defectives, a level of quality that the auto parts industry, among others, have long exceeded, by non-statistical methods. I know of one case of a Toyota supplier that had produced more than 1 million units without a single defective by using Toyota’s Jikotei Kanketsu (JKK) and Change Point Management (CPM).
In the most successful semiconductor companies, yield enhancement involves the combination of the following:
- Knowledge of process physics and chemistry.
- The ability to mine test data on finished circuits.
- Statistical design of experiments.
These skills are almost never found in the same individuals. I believe that Six Sigma Black Belts were intended as a solution to this problem. The idea was to give solid statistical training to 1% of the work force and let them be a resource for the remaining 99%. The Black Belts were not expected to be PhD-level statisticians, but process engineers with just enough knowledge of modern statistics to be effective.
As a metaphor, Black Belt also made sense because there is a parallel between the Six Sigma and martial arts training. Traditional masters in the martial arts of China trained one or two disciples at the Bruce Lee level in a lifetime, just as universities train only a handful of experts in statistical design of experiments who could be effective in electronics manufacturing. One Karate instructor, on the other hand, can train hundreds of Black Belts, just as a Six Sigma program can teach a focused subset of statistical design of experiments to hundreds of engineers.
The effectiveness of Six Sigma is unquestioned in the high technology niche for which it was developed, but the inventors of Six Sigma would not be content with just this application. To market it to other industries with different needs, they removed the technical content and retained only the management parts, with Black Belts and the DMAIC problem-solving model. It was an initial commercial success, with massive adoption by companies like GE and Raytheon. This success, however, did not endure and even GE has now abandoned it. For technical content, some companies are now using the old SPC and calling it Six Sigma, but the most successful repackaging attempt is “Lean Six Sigma,” in which the technical content is entirely from Lean, with only the title of Black Belt remaining for implementers.
The combination of Lean and Six Sigma in “Lean Six Sigma” reminds me of the French recipe for Lark Pâté, which calls for horses and larks in equal numbers: 1 horse to 1 lark. These migrations have caused confusion and an identity crisis among black belts, as reflected, for example, in a long discussion in the Lean Six Sigma Worldwide discussion group on LinkedIn about the meaning of Six Sigma.
I first heard of this approach from an article in Fortune magazine in the early 1980s, describing a secret algorithm called OPT (Optimized Production Technology) developed by an Israeli physicist named Eli Goldratt who did not claim any manufacturing background but did claim to have achieved spectacular results. It was disturbing on multiple levels:
- As I was already aware of at the time, there is more to production technology than just scheduling. To this date, production control remains the focus of TOC, and it is predicated on the fallacy that the physical arrangement of machines and work stations on the production floor does not matter. This is diametrically opposed to the Lean/TPS approach.
- I was developing software for production planning and scheduling in the semiconductor industry. As part of this project, I was studying what the theory of scheduling had to offer, and did not find the notion of a secret algorithm attractive or credible. What I found was that the key challenge was generating schedules that could be executed. Advanced algorithms didn’t work without data at a level of detail and accuracy that was unavailable.
OPT sounded like snake oil, yet Eli Goldratt had managed to get it covered in Fortune! Soon afterwards, I saw a video of a presentation by Goldratt on how accounting was the number one enemy of productivity. His points made sense, and much work has been done since then to improve management accounting. The video was amateurish. Partway through his presentation, you saw a magnet fall off a white board and the speaker himself bend over to pick it up. And he made repeated references to “Murphy’s Law” as if it were more than a joke. The rhetorical style was also unlike any business presentation I had ever heard. While Goldratt was proud of his academic credentials, he spoke like a preacher or a televangelist, which was undoubtedly a marketing decision. Academics who specialized in Operations Research (OR) did not know how to speak to plant managers, but Eli Goldratt did, and many listened. As for televangelists, the real question was whether he was leading them in the right direction. Like many at the time, after reading The Goal, I was willing to suspend disbelief, and, in early 1986, received training as an OPT implementer. My first assignments after that were to participate in implementation projects in two high-volume/low-mix factories for which the approach did not fit.
The first plant made eight models of roller bearings and was laid out as a giant job shop, with rows of identical machines so long that the end vanished in a haze of oil mist, with the corresponding mountains of WIP between operations. The second one was an aluminum foundry centered around 50 diecasting machines working in parallel with multiple machines making the same product. Scheduling was not the key issue in either, but the company had made a corporate decision to deploy OPT in all of its tens of factories, whether it made sense or not. For these reasons, I stepped away from OPT and Goldratt.
Shortly thereafter, his organization stopped marketing OPT, which was not professionally engineered software anyway, to focus on what he called the “thoughtware” behind it, what later became the “Theory of Constraints” (TOC). If phrased as the need for organizations to focus on what is preventing or limiting their ability to reach goals, it is obvious and, at this level of generality, not markedly different from eliminating waste or focusing on “value adding” activities. When you consider specifics, such as production operations, it reduces to an execution methodology called drum-buffer-rope, which still ignores the elephants in the manufacturing room, including how to design factories, lay out production lines, engineer work stations, and apply human resources.
The MIT article is largely dismissive of Agile Manufacturing as “as a patchwork of plausible concepts and methods,” an assessment I agree with. In manufacturing, it is dead, as reflected, for example, by the absence of substantial LinkedIn groups on the subject. Two points, however, still need to be made about Agile:
- The term is still used in software development and project management, with LinkedIn groups on these topics that are of comparable size to those about Lean.
- Roger Nagel’s team at Lehigh borrowed the registered service mark “Agile Enterprise” without attribution from Crispin Vincenti-Brown and Ingersoll Engineers. Their Agile Enterprise was a manufacturing company with production organized in their version of cells, grouped in focused factories, as defined by Wickham Skinner.
Business Process Reengineering (BPR) is also dead, after a burst of popularity in the 1990s. Its concepts and philosophy, however, live on in a kindler, gentler version called “Business Process Management” (BPM). The differences between the two are explained in a blog post by Sweeta Anand. BPR failed in the market because the employees of companies that tried to implement it perceived it as a threat, before its technical shortcomings ever had a chance to appear.
BPR is not about manufacturing but about business in general, and is predicated on the assumption that any business can be structured as a family of processes, characterized entirely by the outputs they generate and the inputs they use. In the BPR perspective, R&D is a process with product designs as output. Manufacturing, on the other hand, is not a process but is instead subsumed under order fulfillment. BPR takes Wickham Skinner’s focused factory or Womack and Jones’s value stream and expands it to all business activities. Each process has its own technical and human resources, and only interact when the output of one process is the input to another. BPR is intended to break down the functional silos that make bureaucracies slow and unresponsive and replace them with process-aligned structures that are focused on their useful outputs.
According to the founding fathers of BPR, Michael Hammer and James Champy, part of the definition of a process is that its output must be “of value to the customer,” but what is meant by customer is ambiguous. Manufacturing is not a process because its completion does not place a product in a customer’s hands but Product Development is, even though it does not do it either. The output of Product Development is used internally, by other processes, like Order Fulfillment, that are only metaphorical customers. One of the principles of TQC is “the next process is the customer,” which means that you should treat the users of your output inside the company as if they were customers who paid for it and had the choice of shopping elsewhere. It is a useful metaphor, but blurs a vital distinction, and, in the case of BPR, creates ambiguity. We could easily argue that Sales is the customer of Manufacturing, and that Manufacturing is therefore a process.
The process model is also not an obvious fit for all the activities of a company. A production line processes input materials into products; an office may likewise process applications into accepted and rejected applications. In these cases, the input-process-output model applies naturally. On the other hand, any activity that can be described as maintenance is usually thought of in terms of its objects rather than input-process-output. It takes work orders as input, processes them by dispatching technicians to the machines, and produces completed work orders as output, but this is an administrative view of the activity. The technicians perceive their job as keeping the equipment up and available, not processing work orders. Likewise, a Master Data Management department executes transactions, but only as a means to the end of keeping a database of specs up to date, and this database is the object around which their work is organized.
In terms of processes, there should not be a Shipping & Receiving department. This function should be distributed among the order fulfillment processes of the different product families. It doesn’t happen because the focus of Shipping & Receiving is the external entities it interacts with, the truckers. If it were distributed, an incoming truck might have to deliver the same item to multiple docks in the same plant. Shipping & Receiving remains centralized because it is a point of contact with world outside the plant.
Technically, BPR is simplistic, but it’s not the reason it failed. What is absent from the BPR literature is any consideration of the people doing the work and what happens to them as a result of reengineering. In manufacturing, Lean improves performance, the company grows, and the people freed up by productivity increases are used to support the growth, and it is essential to the success of the approach that operators not be putting their jobs in jeopardy by participating. BPR, on the other hand, simply reengineered people’s jobs away, which, predictably, resulted in mutinies.
Considering that BPR is essentially equivalent for business in general to converting job-shops to flow-lines in manufacturing, why is it that it can be done without firing people in manufacturing but not necessarily elsewhere? A machinist who knows only one milling machine can practically be cross-trained on lathes, broaching, drilling or grinding machines, or more modern milling machines; an assembler, on different stages of assembly, or on subassembly. But it doesn’t generalize to every business activity. It is a tall order to turn an accountant into a design engineer, or vice versa.
The ineffectiveness of isolated functional “silos” is the object of the 9th of Deming’s 14 points. As we have seen, however, Deming does not prescribe reorganizing along process lines, and it is not always feasible or advisable.
Over lunch at a Lean Forum in Cheboksary, Russia, I was seated with consultants from Russia, Japan, and the US, including Mark Warren. One of the Russians asked me about Lean, and I confesses that, while I was using this label, what I was helping clients with would more aptly be called Baudin Production System, just as Mark’s should be called the Warren Production System, and the same would apply around the table. Indeed, it would be abnormal if, after a couple of decades of doing this kind of work, we didn’t have our own twists on it. The key issue is what clients expect from consultants.
Most of the people who call themselves consultants are really contractors, hired to produce outputs that are needed occasionally, but not often enough to justify hiring employees. The contractors are a temporary extension of the work force, and their knowledge and skills leave with them at the end of their engagements. Then, there is a second category of consultants who are really trainers, hired to help an organization comply with external mandates, such as regulations from the US Food and Drug Administration (FDA) or Federal Aviation Administration (FAA), ISO-9001 certification, or certification as a Lean supplier. These consultants coach clients in the art of passing a particular kind of audit, and are needed as a cost of doing business. Their emphasis is not on improving performance but on providing auditors with what they need to see in order to check-mark the items on their lists. This is the realm of named, formal processes where no deviation from a standard is tolerated. They transfer knowledge, but this knowledge is procedural, and usually has no value beyond ensuring certification.
The last category are consultants who are brought in to help clients improve performance. There, what matters is less the ideas they already have than their ability to come up with new ones, as well as ways to implement them that are feasible in the existing organization. It is similar to solving a Harvard Business School case, with the differences that it is a real situation and that the issues are technical as well as managerial. Top management has identified shortcomings in the company’s operations, feels they must be remedied to remain competitive, but has not found the ability to do it with internal resources. The external consultant brings in a fresh pair of trained eyes, the experience of similar situations, and a kit of tools, either mastered personally or made available through colleagues. The consultant’s generic abilities must then be melded with the specific business and technical knowledge available internally to come up with innovative, ad-hoc solutions. This kind of work requires direct observation of the operations, communication with people at all levels in the client organization, and analysis of the company data. It cannot be done dogmatically. The solutions will vary, as will the means of implementing them.
This is what my colleagues and I do. It cannot be done rigidly or dogmatically. It requires us to be open-minded and evaluate ideas on their own merits from wherever they may come. We operate under a flag or label but, whatever its orthodoxy is, we are always heretics.