What Bill Gurley Saw

Something really interesting happens when you consume a boatload of information: your brain begins to pick out patterns from the mass of stories in your head. This is true regardless of whether it’s books that you’re reading, or podcasts, or even just research programs that you’re executing (I know friends who — on becoming a parent — go deep into child development research; the same thing probably happens for them once they cross a certain threshold of information consumption.)

This is a companion discussion topic for the original entry at https://commoncog.com/what-bill-gurley-saw/

This was super interesting. One takeaway for me was that the study of interesting ideas (Scale, Complexity, Growth) coupled with exposure to real life implications of those ideas (by taking risks as an entrepreneur or investor) helps create tacit knowledge which can lead to outsized returns

Figuring out the right ideas to pick and taking risks without ruin is another matter though. But this was one of the best illustrations of Action Produces Information

Related: Mike Milken on how he came up with the idea for high yield bonds:

In Milken’s case, it was W. Braddock Hickman’s Corporate Bond Quality Investor Experience, which he stumbled on as a Wharton MBA. The book described how the best debt turned out to be credit extended to companies, and the worst debt turned out to be sovereign debt. Milken then worked out the implications of those findings for the rest of his finance career.

I think there’s an interesting aside here on whether it’s possible to gain these sorts of informational edges today, given the penetration of information technology. I think there’s a good case to be made for how you can’t.

That said, it’s surprising how often this pattern crops up when you read about inventors, investors, and people with remarkable careers.

1 Like

I would disagree with that conclusion. Given that the amount of available information is growing massively every year, my hunch is that if anything, the informational edges that one can obtain these days are even bigger. Nobody can come even close to understanding everything in even a small niche, so the variation of what people know and can pattern-match for should be even greater these days.

Sure, much of the newly generated information is not relevant or valuable, so the skills needed to gain informational edges change more from discovery to recognition. It might be less about digging out or stumbling about a gem, and more about sifting through the chaff to find what might be worthwhile to pursue deeper. Developing judgement and taste might be more important than sheer information access.


This is a very fair point.

I stand corrected. I think you’re right — the nature of informational edges changes to favour those with better taste or attention.

I realise there’s another question implied by this post: how does this fit into ‘seek ideas at the right level of abstraction’? The story of Milken and Gurley play directly into our preconceived notions of genius — they discovered some secret insight, and then worked out the implications of those insights over the course of their careers. Does this not violate the argument I made in ‘right level of abstraction’?

I think this is one of those cases where the idea of the ‘narrative fallacy’ actually does matter — it’s easy now to look back and say that Milken’s and Gurley’s big ideas translated to lots of effective real world action. But I think if you take a closer look, you would find that there was a great deal of trial and error involved. Gurley’s initial insight might have been “there are businesses that benefit from increasing returns to scale, and I should learn to identify and invest in those”, but I bet there were lots of failed applications of the idea before he learnt “oh, there is this thing called liquidity quality, and it’s different for every marketplace business, but I can learn to identify companies that have that and management teams that understand that!”

Of course, there’s also the possibility that Gurley was working at the right level of abstraction — Arthur made a number of policy recommendations to governments as a result of his work, and he did talk about the history of industries and certain businesses in his research; Gurley was the tech analyst at Credit Suisse First Boston at the time of his discovery of Arthur’s ideas. One could say that Gurley was simply one level of abstraction below Arthur’s domain.

But I do think there’s a gap between:

  • Brian Arthur’s Increasing Returns (theory) <—> The dynamics of network effects in Internet marketplaces (application)
  • W Braddock Hickman’s Corporate Bond Quality study <—> The actual implementation of high-yield/junk bonds in the 70s, and the various mechanisms pioneered by Milken (application).

The former might be the insight, but there sure is a ton of nuance when it comes to the actual application of the ideas.

I think a lot about Deng Xiaoping’s quote “cross the river by feeling the stones” (摸着石头过河), which he said in relation to China’s shift to a market economy. In Deng’s time, a transition from a communist country with a centrally planned economy to a communist country with a market economy had never been attempted; Deng was trying to express this idea that they knew where they wanted to go, but the actual path to get there was opaque; you had to move by ‘feeling each stone’, step by careful step.

Something similar probably happened with Milken and Gurley. I believe Milken didn’t come up with the idea of using junk bonds to fund hostile takeovers — at least not by himself — his original idea was that he wanted to ‘democratise access to capital’, and many of his deals were with capital-intensive industries like cable and tele-communications. I suspect the game around hostile takeovers was likely proposed(?) by a corporate raider, or perhaps the idea was already in the air in the 70s. Similarly, I’m sure Gurley picked up several foundational ideas when he was involved with eBay in the 90s.

(Note: I couldn’t get hold of a copy of eBoys, which documents the eBay part of Gurley’s story, and I’ve not looked too deeply into Milken’s relationship with the raiders … so I’m not sure of either of these points. I’ll update this if I’m wrong.)

All this is to say that it is easy to make it seem like “get secret idea, win at life” is the whole story; but I think the actual story is a bit more complex, with a lot more trial and error than first meets the eye.

For what it’s worth, I find myself thinking things like “oh, the rise of the creator economy means that I would succeed if I pivoted Commoncog to serve content creators!” and then slapping myself on the wrist and thinking “No Cedric, that’s stupid, that’s the wrong level of abstraction, you should be testing a lower level: are there any content creators who would use a better second brain/read-it-later app? If so, great, maybe the bigger trend is right, but if not, then obviously that trend isn’t playing out for you in the way that you thought it might.”

This sounds stupid now that I’ve written it out, but this thought sequence really happened, and I think it’s the right way to think about such things.


One of the strange dynamics in Silicon Valley is that many investors are operating on a level of abstraction that is different from entrepreneurs. They recognize patterns and see trends, based on exposure to many startup pitches, that operators may not see. They use their theses as filters as they search for startups. But the trend layer of abstraction they operate on is not the right layer to trigger an investment decision. For that, they need more knowledge about the company in question. When I hear you talk about “layers of abstraction”, I tend to think there is a “right action” associate with different layers. By that I mean: at the trend level with ideas like “creator economy”, the right action is to search for ideas but not to pivot in what you build. The trends, if they are real, are most relevant early in the decision making process. Whereas ideas that come from user surveys and testing, in that conversation that makers/entrepreneurs have with potential users, would trigger a “right action” at the level of product development… Most entrepreneurs, because of their love of both making and certainty, jump straight to the building, instead of the “mom test.”

Separately: I’ve read eBoys. It’s a good book. Lots of lessons there. One is that Web Van was more complicated than people think. People make fun of it now, and don’t consider its merits. But it was essentially Instacart. A good idea implemented too soon, with a founder that plowed a lot of his own money into it because of his sincere belief.

Another is David Beirne’s early epiphany about selling to startups. He was a recruiter before he joined Benchmark, and he sold to early-stage tech companies through their VCs. They were the channel, because every VC has lots of portfolio companies and the good ones are a credible source of recommendations to the founders they back. A lot of people new to tech don’t get that.

A third, of course, is the whole story of eBay itself, which was pretty exciting at the time. The chaos of its growth. Its origin as a side project at General Magic, whose story is less known than other SV inflection points.


This is fantastically put. I’ve actually been considering the idea from the opposite direction — that is, you do the ‘mom test’ and not care about any high level trends, because mainstream opinion on what’s a ‘current’ trend probably means that you’re too late to the trend — or at least its effects are too diffuse — for you to take advantage of. What seems to have worked instead is some form of

  1. You discover something that works.
  2. You have no idea why it works but you keep exploiting it.
  3. You eventually realise that it works because of trend X, which you stumble upon in some mainstream publication a few months after you’ve already started exploiting it.
  4. Thereafter, when you tell the story of what you’ve done, you make it sound as if the trend was obvious to you from the beginning, and that your activities are genius-level exploitation of said trend.

(I’m aware that Gurley’s path might be similar to this, as well!)

Dammit, your description is compelling. I’ll need to find myself a copy.


You’re right to point out that there isn’t a single process to get to the good idea, certainly not one that encompasses both investors and operators. I see a couple trajectories:

  • inside-out - this starts with the mom test, or otherwise intimate knowledge of the user. Sometimes the entrepreneur is a proxy for the user. Sometimes the user is their family and friends. Sometimes they simply fall in love with the problem and then solve it. After they have a solution, sometimes their curiosity leads them to theorize about it and situate it among larger trends, which can lead them to other people solving similar problems, whom the entrepreneur can learn from.

  • outside-in - this is what “thesis-based” investors do. Start with the trend, and find the team most likely to succeed within that trend. Sometimes, though, I think entrepreneurs do this early on in the sense that they are “excited about” social networks (think Reid Hoffman or Mark Pincus before they started LI and Zynga), or AI or crypto. Sometimes an operator has a nose for the interesting, and then bathes in that milieu for a while, building an intuition for other people there, until they come upon a problem worth solving.

  • sideways-in - another investor tactic. Find a strong team, and trust them to stumble upon a valuable problem they can solve, which hasn’t been identified yet. You plan on talent making the right pivots.

And these are not mutually exclusive. They are ways to find the right problem, and then the right solution, that an operator can oscillate between.

I have seen both outside-in and inside-out work for entrepreneurs.

inside-out: Some start out with a deep and intimate knowledge of the field, build something and then found a company on it (Jay Krebs et al at Kafka).

outside-in: this is Gurley in a way. Understand some big idea and apply it to a new domain.

sideways-in: See the history of Segment. Or Paypal. Major pivots.

1 Like

I can attest to this: I read Waldrop’s book circa 2000-2001 and didn’t get so far as to consider the dynamics of network effects in internet marketplaces. (I read the book as intro research for possible graduate studies in complexity theory. I enjoyed the book but got a job instead, and didn’t dig much into Arthur’s work.)

Supposing an alternate history where I had reached Gurley’s exact conclusions, I suspect I would have generated different results, and may have made nothing of those conclusions. Context matters a lot, to the point of being decisive in many cases. And I certainly didn’t have Gurley’s context.