My experience with startups extends to starting Gemba Research, a consulting and training company which we merged last year with Kaizen Institute, Gemba Academy, an online lean training venture, and three other business selling lean-related goods and services. With the exception of Gemba Academy, none of the ventures involved information technology in any meaningful way, and even with that one my contribution in that area. We’ve pivoted a few times along the way, probably not enough. It’s possible there are more failed startups in my past that I’ve chosen to forget. The lessons are more dear, and harder to forget.
The past 19.5 years of my career has been focused on the field of lean. It’s with mixed emotion in recent years that I watch lean go fully mainstream (or “terminal” as my friend Jim would say) by extending into all corners of the world, all sectors both public and private. Lean has proven remarkably resilient in spite of its inarticulate naming and a history of suboptimzing around techniques, tools and short-term gains. Lean persists because there are plenty of us still who don’t practice it so well. Lean persists because it approximately works, for the vast majority of organizations and people who give it a good go.
So it was with great interest that I finally made time to read The Lean Startup by Eric Ries which brings the ideas of lean not to a particular sector, but to a particular stage and development cycle of business, to the particular practice of entrepreneurship. I found the book to be a breezy read, well-written and sparing in the use of lean lingo and jargon (kanban, genchi genbutsu), with a genuine and internally consistent spirit of continuous improvement evident throughout.
The main lean ideas involve speeding up the cycle from idea to product to profit by deliberate experimentation and learning. The adaption the OODA learning loop (itself a variant of the PDCA cycle) to form the “build-measure-learn” loop is clever innovation that captures this idea for startups and beyond. In some ways the startup environment is more suited to lean than the mature company, as a greenfield environment with a bias to change.
At the same time, nearly all organizations I have consulted with over the years share the same structural barriers to lean with startups: accounting systems that drive local optimization and build queues, management infrastructure that is not fit to purpose, and failing to systematically value learning. These are all discussed by Ries in the book.
The other notable concept is MVP or Minimum Viable Product. Much like the process and equipment design notion of moonshining or “3P” in lean, the product is released barely ready so that it can be user-tested, improved and rapidly validated or revised. Or as they say at 3M, “Make a little, sell a little”
The term “pivot” seems to have entered the lean lexicon thanks to this book. In startup terms, it means to take feedback of failed experiments and change direction, to find another business model or path to the vision. In lean problem solving terms, it is to try a different countermeasure. There is a certain ambivalence, almost a battle between heard and heart, which can be detected in the author’s discussion of pivoting, hypothesis testing and the use of human judgment in when to persist or change course (pivot). Hypothesis testing may start with an observation, a judgment or even a hunch to develop a theory about what the market needs, followed by testing this against experimental results. Ries correctly points out that human judgment is flawed, but quickly redeems judgment, as “luckily” we can improve our judgment through experimental results. Lean Startup does not go as far as to say that deciding based on data is a must for this learning to be rigorous.
Even though flawed, in the Lean Startup approach sometimes it is OK to pivot based on judgment. This is more startup thinking than lean thinking.
In Ries’ discussion of batching, it is true that small is generally better but not always. One piece flow may be the ideal for discrete production due to ease or problem exposure, balancing work between processes and minimization of in-process stock. However in knowledge work or highly variable service processes, batch size is a question of queuing theory. A supply chain, or even an engineering process, that does not have sufficient WIP will create delays downstream, resulting in suboptimal WIP-building to keep busy, wasteful under-utilization of a critical resource, and other undesirable behaviors. These are wastes caused by variation and overburden, principles that are poorly understood without a good grasp of the data across the entire process. The software development kanban illustration touches very briefly on this, but a deeper discussion of variation and its sourced, and the role of overburden would be beneficial to students of Lean Startup.
The section on 5 Why gave me pause. There are three major problems with Ries’ explanation and recommendations. First, he cites a passage from Taiichi Ohno, explaining that the 5 Why helps companies solve problems and accelerate by allowing “proportional investments” to be made at each level of the 5 Why questioning. This misses the point that the 5 Why drill down is not intended to generate more countermeasures (1 for each of the 5 levels) but in fact it is meant to raise fewer countermeasures at deeper levels in the root cause analysis, ideally just one simple countermeasure at the true root cause. Taking countermeasures or “proportional investments” at each level of the 5 Why analysis merely adds cost in the real world, because attacking the root cause prevents all hither level “whys” to be blocked. If this is not the case, the 5 Why analysis has not been performed properly or at all. The “therefore test” has not been passed, and the 5 Why questions are simply 5 separate causes rather than linked in a logical causal chain.
Second, Ries gives no indication that he understands that “5 Why” is not aimed at finding five actionable causes, but rather a guideline to ask why a minimum of five times, or fewer than five if necessary, merely until the root cause has been identified. The number 5 is a relic of 5W1H and Ohno’s penchant for wordplay. Taiichi Ohno once scolded his student, “We don’t need 5W1H. Just ask “why?” 5 times!” The 5W1H are “what, where, when, who, why and how” used to tell stories, when Ohno only cared about genchi gembutsu, close observation, relentless root cause analysis and a bias for action.
As the diagram below shows, the 5 Why process is better understood when paired with a fishbone (cause and effect) diagram or tree diagram. The first why can lead to several branching causes, which in turn can branch into more. A complex problem may create a tree with dozens of nodes, only a few of which may be the root causes. The “proportional investment”, while sounding like a good idea when limited to only 5 countermeasures, quickly becomes an expensive exercise when applied to the correct application of the 5 Why process.
The third point of concern is Ries’ claim that 5 Why exposes the root cause to be human factors, writing “every seemingly technical problem is a human problem.” This is an incorrect understanding of the 5 Why principle. The golden rule of problem solving is to never find the root cause to be “human error”. Human errors are caused by poorly designed processes, inadequate standards, weak governance, poor hiring practices, etc.
Ultimately these are all processes and systems designed by people, but even then we cannot assign “human error” as the root cause. The reasons why they are designed incorrectly (or not at all) need to be exposed at the incentive, motivational, cognitive or organization structural and system levels. It is at this level of culture that organizations win or lose in the long run. Simply making proportional investments at higher levels in the 5 Why chain is a waste of time that will not prevent problem recurrence for common root causes at these deep levels, except where specific fixes have been made to correct specific flaws. The underlying non-human and system-level root causes remain in the organization and will find their way, like water, inexorably through the cracks in the process.
I encourage anyone curious about the Lean Startup model study alternate sources to gain a better understanding of 5 Why, before practicing it.
Writing “organizations have muscle memory”, Ries recognizes the challenge of creating lasting behavior change in an organization. Yet like so many early lean manufacturing applications, the Lean Startup concept seems overly concerned with tools and techniques to make the work flow and not concerned enough with the leadership elements, the incentives and organization structures required to make lasting change happen. This is understandable as the role of the CTO or tech company CEO is often not organizational development but innovation, thought leadership and marketing. The technology startup in the Silicon Valley garage is notorious in popular consciousness for being the antithesis of a mature, well-structured organization, a fact Ries is well aware. He cautions several times against the drift of large organizations toward bureaucracy and away from agility. Any discussion of lean needs to include the importance of small teams, linked checking of processes and coaching by leaders, pervasive practical problem solving as elements of what is boringly called “daily management”. Daily management is still much neglected within organizations tackling lean from the tools and systems point of view, even in manufacturing, but for greenfield companies and startups, adoption should be quicker.
The concept of “entrepreneur is a job title” is a good one. Taking it a step closer to “Lean” than to “Startup” we can say that “problem solver is a job title”. Innovator, entrepreneur or inventor are certainly sexier job descriptions than problem solver, which can imply mopping up rather than building something new. We can say that invention is nothing more than solving a problem. Innovation is often recombining existing solutions in new ways or for new markets, and entrepreneurs are nothing if not problem solvers, the successful ones at least. The point is, for the majority of people to whom entrepreneurship is not a short-term option, a grasp of the portability of problem solving skills from a traditional organization to a Lean Startup environment, should be empowering.
Ries concludes by humbly submitting his work to rigorous scientific inquiry, asking “Can we use the theory to predict the results of the proposed change?” The concept of a Lean Startup is itself a startup, and Ries shows that he is willing to pivot if necessary. This is an important attitude, because nothing fails like success, and by all appearances the book has done well. The stories told of successful “lean” startups in the book are persuasive, red meat for entrepreneurs eager to learn how to use “continuous innovations to create radically successful businesses” as the cover advertises. Yet the sample size of data is very small. The hypothesis about Lean Startup is more induction than deduction at this point.
Just as many of the exemplar companies in Jim Collins’ bestseller Good to Great turned out to be just lucky, or victims of regression to the mean and not built to last, only time and data will prove or disprove the Lean Startup hypothesis.
There is much we can do to test, challenge and advance ideas in The Lean Startup. A serious application of lean principles to the notion of a startup should start by questioning the very premise. Ries’ definition that “A startup is a human institution designed to create a new product or service under conditions of extreme uncertainty” begs the questions, why extreme uncertainty? What exactly is meant by extreme? What things are uncertain? What is knowable but unknown? What is knowable but ignored in favor of false beliefs? What things are unknowable and represent genuine risks? The key point is that professional management exists to handle these questions, minimize risks, and allow those with the entrepreneurial spirit to focus their creative energies and intuitive abilities on only the “necessary uncertainties”. I challenge the Lean Startup community not to accept “extreme uncertainty” as a given.
In the epilogue Ries asks, “What percentage of all of this waste is preventable?” This begs the larger question of how much of the wasted effort of failed startups can be prevented. The vast majority of entrepreneurs believe their venture will succeed, yet 82% of first time entrepreneurs fail. Fully 80% of people think they are above average, yet this is mathematically impossible. The statistics do not encourage us to dive into startups, and we should listen closely to what data tells us. Surely the Lean Startup methodology can guide a portion of these to success. However there is also the unintended consequence of encouraging unworthy startups to make the leap of faith, or even prolonging the life of startups that do not deserve to be funded, using Lean Startup methods.
Similar to the allures and dangers of Scientific Management which Ries points out, the Lean Startup concept is at a critical juncture: to continue as a popular movement that builds around a business guru or to buckle down, do the homework and shape it into a science. Like Scientific Management, which was essentially an embryonic version of industrial engineering applied to industry a half century before the days of modern management, the Lean Startup methodology approximately works. It could approximately work for years. Nothing fails like success.
If we are believers in reducing waste, a true science of startups should have a rigorous and statistics-based method for filtering and scoping startups before time, money and energy is devoted to them. In a free country, we have the right to pursue business ideas which we firmly believe will succeed even though the odds are clearly against us. We are free to fail and learn. Failed ventures can provide the humus from which sprout the seeds of the next ventures. Encouraging individual initiative is the American way. It is time to learn from a half-century of lean manufacturing and decades of failed startups, to develop a management system for entrepreneurship.