Do We Learn From Our Mistakes? Or Not?
The NYTimes reported a few years back on a Harvard Business School study of venture capital-backed entrepreneurs to test whether or not we learn from our mistakes. The results are confounding to many—including me.
Here’s the story. Several thousand VC-backed companies were studied over 17 years. First-timers had an aggregate success rate of 22% (success meaning going public).
The study is about those trying for a second time. Did the 78% who failed the first time learn from the experience, and do better the second time? Or worse? How did the 22% first-time winners fare—did they get lazy and decline? Or did they somehow do better the second time?
No less an expert than Gordon Moore, sainted ex-leader of Intel and the author of “Moore’s law,” said “You’re more valuable because of the experiences you’ve been through under failures.”
I’m with Gordon. But according to this data, we’re both wrong.
Those who succeeded the first time upped their success rate, to 34%. But those who failed the first time stayed mired in the muck, at 23%. So much for the myth of the gritty, plucky lads who pick themselves up and learn from their failures.
Apparently the data are not the problem: “the data are absolutely clear,” says Paul Gompers, one of the study’s authors. Yet it is still far from clear what the data mean.
As is often the case, data are one thing, and explanation another. Of course, the obvious explanation may be true: people just do not learn from adversity. This seems to be the study’s authors’ view—that the learn-from-failure ethos celebrated in Silicon Valley is really just anecdotal tales over-told.
Then again, maybe we actually do learn more from success than from failure. If so, perhaps that’s because of increased confidence resulting from one win.
Or, maybe only the really good people learn at all. And they can learn from experience alone, whether success or failure.
Or, perhaps these conclusions are only true of a certain type of person, characterized by some cross-cutting characteristic, such as risk tolerance. (Did you know height is correlated with IQ? True: short people score lower on the same IQ tests that tall people take. Of course, if you separate young children from the adults, or use age-normalized tests, the correlation goes away).
Or, to channel a recent 30 Rock storyline, maybe the first time winners are just very good-looking people who are actually horrible, but live in a bubble in which others let them pass. Hey, you never know!
Causal deductions are never fully provable—thanks, Dr. Hume. But progress can be made toward explanations.
So, what do you think’s going on?
And I’ll throw one idea into the ring, borrowed from Karl Popper, who developed the falsifiability theory of meaningfulness. A theory which is highly disprovable, but which remains standing, is superior to a hard-to-disprove theory.
Maybe people who fail have a much greater chance to learn. Why it is that they don’t still seems a mystery to me.
Daniel Kahneman and Amos Tversky had these issues nailed over forty years ago, including the major sources of bias and prejudice that stands in the way of optimal learning — and anyone presuming to function in the field of decision-making had best know their work as deeply as possible. Their seminal 1974 essay, “Decisions Under Uncertainty,” will get you 80% of the way there. Kahneman’s 2011 book, “Thinking, Fast and Slow,” is comprehensive and up-dates very generously for the later scholarship. Michael Lewis’s new best-selling biography of the two of them is entertaining at the personal level but thin going on the substance.
Jim–interesting reviews, but can you clarify which sources of bias you think are at play here? Can you summarize the key concept from the “80%” so I don’ t have to dig up the essay? Thanks!
Charlie–Happy Holidays to you and yours. Thanks for this article, you really nailed it when you say “the data are clear, but their meaning is not.”
I dug up the published version of Gomper’s paper. It highlights the message is not “experience doesn’t help”, but instead “success breeds success”. If you have a good reputation you are more likely to get good opportunities from the VC industry. I guess this wasn’t interesting enough for the NY Times piece.
I wouldn’t read too much into the numbers as the summary table in the paper actually highlights more questions than answers (for example, 90% of the entrepreneurs ran only one company, most serial entrepreneurs are from California, and experienced entrepreneurs are much worse at late-stage companies than first-timers).
Charlie–Happy Holidays to you and yours. Thanks for this article, you really nailed it when you say “the data are clear, but their meaning is not.”
I dug up the published version of Gomper’s paper. It highlights the message is not “experience doesn’t help”, but instead “success breeds success”. If you have a good reputation you are more likely to get good opportunities from the VC industry. I guess this wasn’t interesting enough for the NY Times piece.
I wouldn’t read too much into the numbers as the summary table in the paper actually highlights more questions than answers (for example, 90% of the entrepreneurs ran only one company, most serial entrepreneurs are from California, and experienced entrepreneurs are much worse at late-stage companies than first-timers).
Daniel Kahneman and Amos Tversky had these issues nailed over forty years ago, including the major sources of bias and prejudice that stands in the way of optimal learning — and anyone presuming to function in the field of decision-making had best know their work as deeply as possible. Their seminal 1974 essay, “Decisions Under Uncertainty,” will get you 80% of the way there. Kahneman’s 2011 book, “Thinking, Fast and Slow,” is comprehensive and up-dates very generously for the later scholarship. Michael Lewis’s new best-selling biography of the two of them is entertaining at the personal level but thin going on the substance.
Jim–interesting reviews, but can you clarify which sources of bias you think are at play here? Can you summarize the key concept from the “80%” so I don’ t have to dig up the essay? Thanks!