22 November, 2014
You can’t improve what you don’t measure
If you are responsible for a process that is producing suboptimal results, Continuous Improvement (CI) wants you. CI strives to create a culture around modelling processes and improving them using disciplined experiments.
CI encompasses many ideas that should be widely (perhaps wildly) beneficial in most business contexts:
Share responsibility for improvement with each person in the org.
Always keep the customer in mind.
Understand the value produced at each stage of your production processes.
Focus on improving the processes more than hitting a quota.
Improve processes by eliminating waste.
This article explores the nature of CI by reviewing its history. In the process, I hope you’ll get a sense of why it’s a big deal.
CI didn’t always look like it does today. It began simply in the 1930s as statistical methods were applied to business practices. It turns out, though, that CI is generic enough it can improve its own process with experimentation. So, the form and reach of CI has evolved over the past eight decades.
Science hit industry in a big way at the turn of the 20th century through a machine shop foreman. Around 1880, Frederick Taylor was unhappy with his workers’ lackadaisical pace and set out to calculate just how much time was being wasted. He broke up their work into tasks, timing each step, and in so doing was able to determine a better set of steps to increase productivity. You can imagine how his workers responded to this, so Taylor developed a system of introducing his observationally-developed process changes that became known as Scientific Management, or sometimes, Taylorism. His system proscribed an unprecedented level of management involvement in workers’ daily routines. After discovering the "One Best Way" to perform any task, management selects workers most suitable to performing those tasks and then trains them under constant supervision. Increases in productivity from applying these methods were remarkable. Unfortunately, the promise that workers would make more money, work more sustainably, take more pride in their productivity, and ultimately be more fulfilled was seldom obtained. In proscribing “One Best Way” for every task, Taylorism was not only patronizing, it also failed to leave room for process innovation.
The Plan-Do-Study-Act (PDSA) cycle was an early application of statistical methods to business management. It was an adaptation by W. Edward Deming of earlier ideas by the statistician Walter Shewhart. Deming presented the PDSA cycle in Japan after World War II, helping spur a resurgence in Japanese manufacturing and establishing him as the honorary "father" of CI. The PDSA model’s conceptual breadth enabled it to evolve alongside developments in CI, so it is still used today. Think of PDSA simply as "experimenting to learn". You look at your current processes and think “Hey, I wonder if this change would make things better…” (Plan), then you try it out on a small scale (Do) and see if it worked like you hypothesized (Study). If it did, then you implement the change (Act). This is essentially the scientific method as a counterpoint to management by intuition.
While the US discarded scientific and mathematical methods of production in the wake of WWII, Japan adopted them as a core part of their post-war recovery. Toyota, in particular, went all-in resulting in the development of Lean Manufacturing (Lean).
Lean expects line-workers to focus on eliminating waste (muda) from their local processes while managers fill a coaching role. By creating a shared culture of Kaizen (Japanese for "improvement"), the many small improvements by line-workers bubble up to organization-wide efficiencies.
"All we are doing is looking at the timeline from the moment a customer gives us an order to the point when we collect the cash. And we are reducing that timeline by removing the non value-added wastes."
Lean is primarily a management system for "the absolute elimination of waste." Waste being defined as:
Costs of quality: scrap, rework and inspection
To readily identify waste, Lean prescribes a focus on flow. A primary tool for this purpose is the Kanban board consisting of each work item (a product or other unit of value) represented on a small card and moved progressively through columns representing stages of the production process. Waste is readily seen if cards pile up or stagnate anywhere on the board. Placing a limit on the number of cards in any given column (a Work-in-Process, or WIP Limit) has the beneficial effect of encouraging all members of the production team to address the blocking issue or source of waste to prevent the production line from grinding to a halt.
Waste may also be called-out in-the-moment by any team member or by automated systems. Some sort of signboard or signal (andon) is used to immediately stop production until the waste is attended to.
To get to the root cause of identified waste, Lean suggests asking the question "Why?" five times. While keeping in mind the answer may involve identifying problems and making changes in one of six areas:
The people involved
The methods currently employed (e.g., policies, procedures, rules, regulations and laws)
Measurements being taken, or
The environment of the process (e.g., location, time, temperature, and culture)
Through CI, Japan produced high-quality products that dominated electronics and auto industries. It took until the 1980s for American businesses to begin implementing their own version of CI.
Keen to understand Toyota’s Lean success, General Motors entered into a joint venture in 1984. Together, they re-opened a GM factory in Fremont, CA named NUMMI. NUMMI had been closed a couple years earlier and had the reputation of having the worst workforce in the US auto industry. Newly supplied with Lean processes and training, the same workers were soon producing automobiles with quality on par with their Japanese rivals.
In telecommunications and consumer electronics, Motorola responded to Japan’s industry dominance with their own Six Sigma (6σ) program. 6σ focuses on rigorously defining and standardizing processes to achieve repeatability so precise that only 3 or 4 defects occur out of a million. To attain this, Six Sigma prescribes the Define-Measure-Analyze-Improve-Control (DMAIC) process. It is pretty much PDSA with a heavily statistical focus on controlling process outputs. 6σ practitioners earn belts like Karate students, so if you progress to the rank of Black Belt, you pretty much are doing process improvement coaching all day, every day.
Not to be outdone, the US Navy developed their own version of Lean in the late 80s, including an all-in approach to CI with the moniker Total Quality Management (TQM). In TQM, the entire organization is "All Hands on Deck". This means every person in the org is expected to have a clear understanding of the org’s Vision, Mission, objectives and critical processes and adopt continuous quality improvement as a core value. In TQM, the org adopts quality improvement as one of the primary (if not the primary) activities of the organization as a whole. It’s hard core.
Standards bodies didn’t want to be left out. In 1987, the International Standards Organization modified existing quality specs for military suppliers to create ISO-9000. In 2000, it underwent a major rewrite to focus on process quality rather than simply end-product quality, and in so doing essentially became a CI process. Manufacturers want quality from their suppliers. This has driven the practice of favoring, or requiring, suppliers to have ISO-9000 certification. Certifications have doubled over the past decade to over 1 million globally. Chinese suppliers now account for about a quarter of all ISO-9000 certifications, with more than 10 times the number of certifications by US companies. The total number of US certifications is currently dropping. This is perhaps because complying with a standards body is not necessarily the same thing as improving an actual process; and the burden of compliance documentation is itself a source of waste.
TQM, too, hit the peak of its management hype-cycle in the 90s and has experienced slower growth since. In fact, many of the companies most identified with CI, including Toyota, Motorola, 3M and GE are now struggling against stiff competition and cutting back on some of the more formal aspects of CI. The manufacturing space may have reached a certain CI saturation point. However, CI concepts ubiquitous in manufacturing have been recently infecting other industries such as service, healthcare, education and IT.
The turn of the millenium saw Lean come to software development in the form of, wait for it, Lean Software Development (LSD). The core focus remained the same, with the primary adaptations being (1) the form of waste and (2) what constitutes the start and end of the value stream. In software development, the process can be thought of as beginning when a feature or product request is made and ending with the deployed implementation.
The 8 principles of LSD (currently) are:
Optimise The Whole (purpose-driven, work on entire value stream)
Focus on Customers
Energize Workers (purpose, challenge, responsibility)
Eliminate Waste (small WIP, Useless features)
Learn First (decide at the last possible moment, learn quickly, respond rapidly)
Deliver Fast (optimize for flow, MVP)
Build Quality In (TDD, automated testing, ci, expect no defects)
Keep Getting Better (attack small failures, test and learn on small things)
Continuous Improvement is an amazing capability for a process-oriented organization. But as with all things, it has it’s limitations. Let’s not get all "You can’t improve what you don’t measure" about it while invoking the management genius of W. Edward Deming or Peter Drucker. Not only because neither Deming nor Drucker ever said it, but because they wouldn’t agree with it.
Deming recognized that managers need to manage even when figures are unattainable. In fact, he said "3% of the problems have figures, 97% of the problems do not." Unlike Deming, Taylor had immense faith that every task had a particular optimal solution and that “scientific” measurement would tease it out. Science has since shown Taylor to be naive. Chaos theory, information theory, control theory and other fields conclusively demonstrate that most human systems are not simply complicated, but are indeed chaotic. Unfortunately, legions of managers continue to mirror Taylor’s lack of humility through over-application of simplistic operations management theory.
Drucker recognized another limitation of CI. Namely, that much of the improvement of an organization comes not as figures, but in the form of relationships within the organization: "Your first role . . . is the personal one. It is the relationship with people, the development of mutual confidence, the identification of people, the creation of a community. This is something only you can do. It cannot be measured or easily defined. But it is not only a key function. It is one only you can perform."
Furthermore, as any statistician will tell you, creating models is a tricky business. Not only can you have limited data, wrong data, irrelevant data, underfit the data or overfit the data, the most fundamental problem is that relying on statistics gives you no absolute guarantee that your model may, at some point, just stop working. Deming said, "Every theory is correct in its own world, but the problem is that the theory may not make contact with this world."
And in the case of quickly changing systems where inputs, requirements and technologies are ever-changing, CI may never even get off the ground.
The greatest limitation, perhaps even danger, I see with CI is that it has the potential to lull an organization into a self-referential sleep. Deming called it out: "It is a mistake to assume that if everybody does his job, it will be all right. The whole system may be in trouble." A system can be optimized locally and yet be a complete waste globally.
Continuous Improvement, at it’s heart is simply about adopting a disciplined scientific approach to modelling a business process and driving out inefficiencies. Its concepts, experimentally developed for nearly a century, provide a rich set of tools to improve your work (mainly on speed and quality, but also team satisfaction). Just don’t go too crazy out of the gate.
You can get started today with any service or production process you use to deliver value to customers. Break that process into steps, and start looking for waste that keeps any step from moving as smoothly as it can. When you find waste, do a small experiment to see if you can eliminate some or all of it. If you can, then document the new process.
"Manage the cause, not the result."
Once you start focusing on the process rather than the outcome, you’ll be on your way with Continuous Improvement. There are a thousands of books to go from there. Amazon will help you find the most popular ones. Also, companies tend to geek out about their CI processes; many of them run tours. A great CI tour in the Seattle area can be had at Kaas Tailored.
2015 Paul Payne