Have you ever wondered why the best venture capitalists keep winning outsized returns? Why do the best agents keep discovering new talent? Why do the best universities keep minting Nobel Laureates and CEOs?
Most people attribute this continued success to their being “the best.” That somehow their innate abilities, processes, and general brilliance enable them to succeed over and over again. But there’s a much simpler and more powerful answer — they have selection bias on their side. In today’s Letter we’re going to talk about this inexorable force and how you can build systems that get it on your side.
Selection bias is the effect on outcomes when the group being studied is not random, but rather has passed a series of selection gates. More specifically, the winners we listed above have carefully cultivated a sub-class of selection bias known as sampling bias. Sampling bias is the practice of operating on a non-random subset of a population.
Is the set of deals that Kleiner Perkins sees consistently higher in average quality than the average quality of deals your local angel group sees? Yes. Are the set of actors WME has the option to sign higher in average quality, and working harder to be noticed, than those available to the brand new talent agency with no track record? Yes. Are the high school students applying to MIT already more likely than those applying to the bottom 95% of colleges to be Nobel Laureates? Yes.
When you start thinking about selection bias, you see it everywhere. Sports team performance, mutual fund growth (if not performance), medical study results – you won’t be able to avoid it. And if you’re not careful, that can be disheartening. After all, if everyone else is playing with a stacked deck, how can you win?
Build Your Own Loaded Deck
Let’s think about intentional selection bias in terms of poker.
You’re playing a very odd game of heads-up Texas hold’em where you and your opponent are being dealt out of different decks. You are brand new to the game, with no reputation. Your opponent is Phil Ivey, who just won the World Series of Poker. The cards are sentient, and they want to get dealt to Phil — after all, cards that get dealt to Phil are known to be played in winning hands! So they are lining up at his deck, and he can choose just the face cards and Aces (and 10s and 9s if we’re being mathematically correct here).
Meanwhile, all you have access to are the cards that got locked out of Phil’s deck when the number reached 52. Now you’re trying to play against Phil, where his deck is full of high cards and you’ve got 8s and below. How could you possibly win? Then, think about the next game. Phil has now won two in a row — do you think the selection bias is going to grow stronger or weaker for his next game? Exactly. It’s a virtuous cycle for Phil and it’s a disaster for you.
Now let’s use an example that doesn’t rely on self-aware playing cards (shudder). You are the best entrepreneur in the world. You have had multiple exits and have just settled on your next big idea. You would like to get venture funding, not just for the cash (you are a winner, after all), but for the team and external validation. Who are your first 5 calls? I’ll bet you the list looks a lot like Kleiner Perkins, Sequoia, Benchmark, Chris Sacca, and Founders Fund. You know who I bet’s not on your list? Random family offices, angel networks, and second-tier VCs.
It’s not just that the big VC and PE shops experience positive selection bias — it’s that everyone else experiences negative selection bias as a result. That’s why new firms that try to copy their strategies fail. The same approach that works for KKR by definition cannot work for a startup PE fund because they are playing with a fundamentally different deck of cards. KKR’s rate-limiting factor is throughput. They have so many well-qualified deals that they’re trying to optimize for returns over time.
Your rate-limiting factor as an upstart is totally different — diligence. You need to comb through an ugly subset, like Billy Beane in Moneyball, to find the team of players that everyone else has ignored. You must do this while knowing that every deal you see has been passed on by organizations with more stringent selection criteria than you. The set you see has been reverse-filtered: with every step in the process, it contains, on average, more feces and less water.
Medical science and Private Equity have known about and, respectively, avoided and courted selection bias for decades. Have you thought about how it impacts you? Are your marketing efforts giving you your choice of a compelling subset of opportunities or do you find yourself acting on a good chunk of what walks in the door? Do you get excited by the Startup/Crypto/Cannabis deal flow you see or do you recognize that it’s a negatively selected subset that is, on the whole, less likely to win?
Three Ways to Get Selection Bias on Your Side
- Downselect further. This seems counter-intuitive. Why would I reduce my optionality in the aim of bettering my odds? Well, when you downselect all the way to an esoteric field, you can dominate the conversation and head opportunities off at the pass before they head over to generalist players. If you believe computational microbiology is exciting enough to invest in, why not start a fund that just focuses on it? Show up to every industry event, hire industry icons under your banner, and you’ll become the player of record far faster than you would in a less insular world. Create a situation where you are so dominant in your sub-domain that someone in your field going to a generalist investor instead of you raises red flags.
- Publicize your wins, not your fundamentals. Funds and startups both make the error of trying to be known for who they are as opposed to what they’ve done. Newsflash, folks, there has never been a VC who hasn’t promised to “be helpful,” and there has never been a startup that wasn’t working on something “innovative” or “disruptive.” Instead of talking about what you want to be, aggressively publicize objective evidence of what you ARE. Is over 50% of your fund focused on deep tech? Write it up. Is your technology beating benchmarks by 5x? Write it up. Select for people who pay attention to objective measures.
- Pre-emptively fire underperforming subsets. This is the most powerful tip, and the one that’s approached with the most apprehension because it portends such perceived permanence. As you’re designing your systems, intentionally exclude the subsets of the population you believe will underperform. You may miss a few diamonds in the rough, which can be scary, but the average quality of your set will be higher. As a startup founder, if you believe that older consumers will be the primary users, don’t be afraid to rag on millennials in your marketing (it’s easy, and fun!). You’re firing a huge subset, but if that increases your targeted likability to a more appealing subset, it’s a win. As a Real Estate investor, maybe you have data to suggest that the southeastern US is the only place new Assisted Living facilities should be built in the next five years. Yes, you might disqualify a few great deals in Arizona, but by creating a de facto focus and firing the distractors, you’re increasing your chance of getting selection bias to work for you, instead of against you.
