We Have Learned Nothing

Click here* to subscribe to print for your office or home.

Any method for building a startup, once widely known, causes founders to converge on the same answers. If everyone follows the same bestselling startup techniques, everyone ends up building the same company and, with no differentiation, most of those companies fail. The truth is, anytime someone insists on a method for how to build a successful startup, you should do something different. The paradox is self-evident, once you think about it, but it contains the seed for how to move forward.

Before the wave of New Punditry began 25 years ago, the body of startup advice it displaced was, admittedly, worse than useless. It consisted of a naïve amalgam of Fortune 500 corporate strategy and small-business tactics, of five-year plans and day-to-day blocking and tackling. But for high-growth-potential startups, long-range planning is worthless. The future is unknowable, and focusing on daily operations leaves founders exposed to faster-moving competitors. The old advice was built for a world of incremental improvement, not radical uncertainty.

The New Punditry’s advice was, instead, intuitively rational, apparently well-argued, and offered founders a step-by-step process for building a business amid real uncertainty. Steve Blank’s customer development method in The Four Steps to the Epiphany (2005), for example, taught founders to treat their business idea as a set of falsifiable hypotheses: get out of the building, interview potential customers, and validate or kill your assumptions before writing any code. Eric Ries’ The Lean Startup (2011) built on this with the Build-Measure-Learn loop: Launch a minimum viable product, measure real user behavior, and iterate rapidly rather than waste time perfecting a product no one wants. Osterwalder’s Business Model Canvas (2008) gave founders a tool to map the nine key components of a business model and pivot when something isn’t working. Design thinking, popularized by IDEO and Stanford’s d.school, emphasized empathy with end users and rapid prototyping to surface problems early. Saras Sarasvathy’s Effectuation Theory prescribed starting with a founder’s own skills and network rather than reverse-engineering a plan to meet a distant goal.

These pundits were consciously trying to build a science of entrepreneurial success. By 2012, Blank said that the National Science Foundation was calling his customer development framework “the scientific method for entrepreneurship,” and claimed that “we now know how to make startups fail less.”[1] The official Lean Startup website claims that “The Lean Startup provides a scientific approach to creating and managing startups,” and the back cover of his book quotes Tim Brown, CEO of IDEO, saying Ries “proposes a scientific process that can be learnt and replicated.” Meanwhile, Osterwalder claimed in his PhD thesis that his Business Model Canvas is rooted in design science (the precursor to Design Thinking).

Academics in entrepreneurship departments also study startups, but their science is closer to anthropology: describing the culture of founders and the practices of startups in an attempt to understand them. The New Pundits had a more practical vision, the one that the natural philosopher Robert Boyle articulated at the very dawn of modern science: “I shall not dare to think myself a true Naturalist till my skill can make my garden yield better herbs and flowers.”[2] A science should seek underlying truth, in other words, but it should also work.

Whether it works or not is, of course, what determines whether it deserves to be called a science. And if there’s one thing we know about startup punditry, it’s that it hasn’t worked.

In science, we discover whether something works by running experiments. When Einstein’s theory of relativity was gaining acceptance, other physicists devoted time and money to devising experiments that would test whether it made accurate predictions. We all learned in grade school that the scientific method *is *science.

Yet through some flaw in our nature, we also tend to resist the idea that this is how truth is found. Our head expects evidence, but our heart demands to be told a story. There is a venerable philosophical position—wonderfully examined in Steven Shapin and Simon Schaffer’s* Leviathan and the Air Pump *(1985)—that observation cannot give us truth, that we can only find real truth by deriving it through logical principles from other things we know to be true, i.e., from first principles. And while this is the standard in mathematics, in any area with slightly noisier data or a less firm axiomatic base, it can lead to appealing nonsense.

Until the 16th century, doctors used the work of the second-century Greek physician Galen to treat patients. Galen believed sickness was caused by an imbalance of the four bodily humors—blood, phlegm, yellow bile, and black bile—and recommended treatments like bloodletting, purging, and applying heated cups to restore balance. Doctors followed these treatments for more than a millennium, not because they worked, but because the intellectual authority of the ancients seemed to dwarf the value of mere contemporary observation. But around 1500, the Swiss physician Paracelsus noticed that Galenic treatments did not actually make patients better, and that some treatments—like mercury for syphilis—worked even though they made no sense within humoral theory. Paracelsus began to advocate listening to evidence rather than deferring to the authority of the long dead: “The patients are your textbook, the sickbed is your study.” In 1527, he even staged a public burning of Galen’s work. His vision took centuries to take hold—nearly 300 years later, George Washington died after an aggressive bloodletting—because people are more inclined to believe neat and simple stories like Galen’s than to confront messy and complex reality.

Paracelsus started with what worked and followed that to why. First-principles thinkers start with a hypothesized “why” and then insist it works, regardless of the results. Are our modern entrepreneurship thinkers more like Paracelsus, driven by evidence? Or more like Galen, sustained by the elegance of their own story? In the name of science, let’s look at the evidence.

Here is the official government data on U.S. startup survival.[3] Each line shows the survival likelihood of companies started in a given year. The first line tracks one-year survival, the second line two-year survival, and so on. What the chart shows is that between 1995 and the present, the percentage of companies surviving for one year is essentially unchanged. The same is true at two years, five years, and 10 years.

The New Pundits have been around long enough, and are widely known enough, that their relevant books have collectively sold millions of copies and are taught in virtually all university entrepreneurship courses.[4] If they worked, it would show up in the statistics. Instead, there has been zero systematic progress over the past 30 years in making startups more likely to survive.

The government data counts all U.S. startups, including restaurants, dry cleaners, law firms, and landscapers—not just the high-growth-potential tech startups that VCs fund and the press covers. The startup pundits do not claim their methods apply only to Silicon Valley-type companies, but the techniques are most often tailored to the kind of radical uncertainty a founder would generally accept only if there were a potentially large payout down the road. So take a more targeted measure: the percentage of U.S. VC-funded startups that raised an initial round of capital and later went on to raise more. Given how venture capital works, we can safely assume that a large majority of the companies that failed to raise a subsequent round did not survive.

The solid line is the raw data; the dashed line adjusts for recently seeded companies that may still raise their S