In Venture Capital, we have lots of rules of thumb for assessing entrepreneurs. Some such rules are:
- Invest in guys who are already rich, because they have fewer distortions in their motivations.
- Invest in guys who aren’t already rich, because they’re hungry.
- Invest in guys who have put their own wealth at risk, because they have “skin in the game.”
- Invest in guys who have raised outside capital from credible investors.
- You gotta outsource the tech stuff — it’s too expensive.
- You never outsource the tech stuff — it’s too important.
- Entrepreneurs should not come from a big company (because they have the wrong culture).
- Entrepreneurs should come from a big company under a big-ego boss (because they want to make their own name out from the shadow of the big cheese).
- Only young guys get this new (social networking, Web 2.0, whatever) stuff.
- Only seasoned old guys will make any money.
When you’ve got guidelines like this, many of which are quoted with nary a trace of irony by the practitioners of our art, it’s fairly clear that you actually don’t have any useful guidelines.
Except one: invest in guys who’ve done it before.
The only person you know has a shot at creating a company with an exit value over $100 M (which is about the minimum exit value that most VCs would admit to wanting in any case, so I’ll use this as the threshold for my definition of a “star entrepreneur”) is the guy who’s already done it. Everybody else hasn’t yet proved it.
However, we have a major signal-to-noise problem here. When you look at an entrepreneur, even if he’s a star, you don’t know to which causes and in what proportions to attribute that stardom. Certainly, the quality E of star entrepreneurship could be the cause. But it could also be quality L (for luck). With only one exit, you just don’t know. But with no exits, you know even less.
If you were a statistician, you might try doing something with a Chernoff bound or a Z-test to figure this out. But a few things confound this approach. The main problem is that your number of trials is usually one. You can’t just keep tossing the coin to see if it comes up heads 51% of the time; you have to somehow guess the bias after only one flip (no pun intended!).
Another big source of error is that successful exit events are fairly rare. Most venture funded companies don’t have successful exits. (The old chestnut about VC is that a fund makes ten investments; five fail, 3 or 4 return 1x capital, and 1 or 2 make a 10x return. That’s hardly accurate, but it’s a good enough schematic for understanding the rarity of successful exits.) So if the probability of any given entrepreneur having a successful exit is 10%, then even a “rock star” who is five times better than the average is still only even money to exit big.
So, we have lots of both Type I and Type II errors. Finally, successive startups are not independent trials. That is, unlike rolling the dice, each startup you do is affected by the previous ones (both the experience of doing them and the ultimate outcome).
Even when an entrepreneur has been successful in the past, we don’t know if he’s likely to be successful again (though we can say, on average, he’s more likely to succeed than a newbie). But often, we fail to reject the guys who “hit the lottery” with their previous success. And even more often (almost by definition), we end up rejecting highly promising entrepreneurs who haven’t yet had a home run.
Yet, somehow, VCs continue to invest, and the returns (compressed due in large part to excess capital seeking a home in alternative assets / private equity) have continued to be good (if less princely than in the early years of VC). So VCs aren’t just monkeys throwing darts; we do have some discrimination ability.
Another chestnut in the VC industry is that “it takes $X million (e.g. $30 million) to make a general partner.” That is, to get a venture investor to the level of a seasoned senior investor, he needs to have led $X million in deals. This amount of deal flow (or, as it’s sometimes told, losses) is required to build up the black box of intuitions, gut feelings, sixth sense, etc. that a good general partner should have.
To me, this is a bunch of horse hooey. Yes, any good seasoned professional in any field will have some pre-rational judgment abilities that appear to be a “black box” — but these should only come into play at the margin. The core of any professional discipline must be reducible to a teachable, coherent syllabus. Take, for example, Malcom Gladwell’s example in Blink of the use of the Goldman algorithm for diagnosing acute myocardial infarction (heart attack). Essentially, the algorithm kicks the ass of expensive cardiologists and other “professionals” at doing one narrow thing, which is telling whether a heart attack is happening.
Now, the human body is a system, replicated and observed over billions of instances, with trillions of dollars cumulatively spent on measuring and exploring it. It provides feedback continuously on a second-by-second basis. And a heart attack is a fairly catastrophic and disruptive event — the death of part of the most important muscle. So the fact that a relatively simple algorithm can discriminate that event with great specificity is perhaps unremarkable.
But what is remarkable is that it took until the mid-eighties to promulgate the Goldman algorithm, and even today it’s not the gold standard. (Hey, at least medicine has the Goldman algorithm: in VC, nobody yet has validated such an approach.)
Doctors and VCs both have acute cases of what I call “special snowflake syndrome:” both groups tend to believe that they have special, irreducible talents and skills at doing whatever they do, and that they could never be replaced by a dumb machine. It is no coincidence that special snowflake syndrome tends to strike those in high-income jobs; folks who’ve seen automation or offshoring put downward pressure on their wages tend not to subscribe to this conceit. We also see special snowflake syndrome in industries where there are relatively high barriers to entry, such as regulatory (medicine) or timing / liquidity (10 year partnership agreements in VC).
In both cases, special snowflake syndrome will inevitably lead to heartbreak as people without the illusion of snowflakeness find and implement things like the Goldman algorithm (and the coming, immodestly named Lucas algorithm for VC). Unfortunately for the doctors, they won’t be capturing the economic surplus that results — it will be the middlemen in the hospitals and insurance behemoths that soak up the savings. (Doctors: save yourselves now, by demanding economic ownership of your patients’ well-being, the only humane and just way to allocate costs and risks in your profession!)
Fortunately for VCs, the implementation within a partnership of the Lucas algorithm will enhance that firm’s ability to identify and make successful investments with greater certainty and less manpower. And since VCs, through the carried interest portion, are compensated on financial performance, this will ultimately benefit the adopters. Yes, there will be some Schumpeterian woe for the old-school hangers-on as the snowflakes of their egos are melted in the sunshine of the new day that will be ushered in. And yes, a certain few of the old-school VCs — the ones with really great black boxes and/or Rolodexes — will continue to enjoy their maverick road gambler reputations. But by and by, rationality is coming to our market. Our black boxes will help us make decisions at the margin, but our algorithms will drive our core activities.
Does anybody want to work
on the algorithm with me? (I’m willing to hyphenate the name of the algorithm.) Drop me a line or give me a call at Voyager, +1 206-438-1822. (There will, of course, be several variations for different industries, geographies, stages, and firm preferences — so much so that each firm will likely need its own implementation — which is why I’m not too concerned about giving away competitive advantage by discussing with others in the industry.)