Skip to content

Marc Andreessen on Scar Tissue, Trying and Failing in the Probabilistic Domain

This was a good interview on Farnham Street’s The Knowledge Project – listen here.

Marc Andreessen: It’s the most natural thing in the world to try to learn from experience. We all went through this: touch the hot stove, it burns, and we don’t touch the hot stove again. And that happens at age three or whatever, and then from on basically… your trajectory through life is like touching hot stove after hot stove after hot stove, learning don’t do that, don’t do that, don’t do that…

… You learn all these things… and what a lot of people learn is, they work at a startup that doesn’t work and so the lesson is “Don’t work at startups.” Or… a lot of people went through in the last fifty years this thing where they thought they had lifetime employment at some big company and then they got laid off at some point, and the lesson is “Don’t work at big companies, because you can’t trust them and they don’t care about you and they’ll lay you off.”

And it’s the most natural thing in the world to learn from these mistakes, and so what happens is… as you age you just naturally build up all this scar tissue and you have all these cautionary lessons in your head all the time. And therefore as a consequence your aperture of what you’re willing to explore as you get older just shrinks. And it’s an incredibly natural thing to have happen.

… My day job is an incredible teacher that that’s a bad idea. What happens in my day job is – I’ve now been doing startups as an entrepreneur for fifteen years, now as a VC coming up for fifteen years – and basically what you see if you swim in the pond for long enough… you see example after example of startups that tried something and failed. And you draw the experience from that, “Okay, well we know that’s never going to work,” because it failed… And then you just see example after example of what happens next which is that some kid shows up with that same idea five or ten or fifteen or twenty years later, and they take another swing at it, and it becomes a gigantic success.

… So Webvan doesn’t work, but today you’ve got these giant successful companies in food delivery that are huge successes, so that idea absolutely works.

Shane Parish: The wrong lesson to extract from that experience is “This idea will never work.” How is that we extract the right lesson from an experience?

Marc: I think you want to really understand the domain in which you’re operating. The domain of startups is… a probabilistic domain… For top-end venture-backed startups about half the companies in any given period work, and generate positive returns and about half of them don’t. So we start out every new fund knowing that half of the companies that we invest in are probably going to fail… They’re going to fail for a bunch of reasons. A lot of the failures are idiosyncratic – the team breaks up or whatever, or something specific to the startup – but in a lot of cases it literally is that they’re going to fail because they’re flat-out too early… it’s going to be the Webvan of its time…

You’re thinking probabilistically. We’re not trying to score 100 on the exam. And this is maybe a big difference between the education system and the real world: in the education system we’re supposed to get 100 on every test. It’s a deterministic realm in which there is a way to achieve perfection and we’re going for the 4.0 grade average.

In “real life” – which is starting businesses, or writing books, or composing music, or playing basketball or playing poker, or basically doing anything interesting, we’re in the probabilistic domain. And in the probabilistic domain we’re not going to score 100 percent… In most of the probabilistic domains, world-class is like 60 percent. I don’t know what the numbers are for basketball, but if you look at Kobe Bryant or LeBron’s lifetime record, they’ve scored all these baskets. And… the ranking of the basketball players who have missed the baskets… it’s the same people.

What Annie Duke says… in Thinking in Bets… if [people] are in a deterministic domain and they get less than 100 percent, then they can self-critique and they can say “Why didn’t I get 100 percent? I should have got 100 percent, and I did something wrong.” And that is a correct analysis. You could have got 100 percent on that test, you probably didn’t study hard enough, you didn’t read all the material or whatever.

In a probabilistic domain, “It didn’t work.” Okay. It’s again this scar tissue. I do what she calls resulting, now I need an explanation for the result. And the explanation for the result is, I should have done this or that, I should have bluffed or not bluffed, or not started this company, or waited five years. And she’s like, “Look, it’s a probabilistic domain. Some percentage of them are simply going to fail. If you try to explain what fundamentally the result of randomness through a narrative, you’re basically fooling yourself. Ironically you’re making yourself miserable in a counter productive way. And the right thing to do, in poker, is to play the next hand of poker the exact same way you played the previous hand because you’re operating in this domain of probability. The right thing to do if you’re Kobe Bryant is to shoot the next basket. The right thing to do as an entrepreneur is to start the next company… even in a space where you’ve already tried and failed before.

And there’s a sort of point a lot of people reach where they’re just done with new ideas. It’s a general version of the problem. Like, “I got through the PC, I got through the internet, I got Facebook… but like, TikTok? Really? It’s too much.” It’s just like, at this point the world is getting ridiculous, and I don’t want to process this anymore. And almost everybody hits that point at some point.

The bad news is you’re going to make all these mistakes. The good news is you don’t have to score 100 percent. And not having to score 100 percent means that you can take chances. Success or failure is not at the unit of an individual investment. Success or failure is at the level of a portfolio… When you liberate yourself away from Ï have to get this decision right,” to “I have to execute a strategy that is going to result in a series of ten bets, and then I’m going to be evaluated not by any individual one – and by the way, I’m not even going to be evaluated by the average outcome, I’m going to get evaluated by the whole performance of the basket. And if one of those bets results in 99 percent of the profit of the basket, that is a perfectly fine outcome [if it’s a win overall].

Not all domains are like this, like the person running a nuclear power plant doesn’t get to live in this world. But anybody doing anything creative does live in this world. And I think that creative artists of all kind basically live in this world and the really good ones are unafraid to take bets because they know that this is the domain they operate in.

Marc Andreessen, interviewed by Shane Parish in The Knowledge Project Episode 129

See also:

Kareem Abdul-Jabbar on scoring and missing
Mass production: quality as a function of quantity

I'd love to hear your thoughts and recommended resources...