How to Back the Right Startup: Picking in a Power-Law World

Here is the number that should be taped inside every corporate investor’s laptop lid. When Benedict Evans got access to Horsley Bridge’s data — a fund-of-funds with visibility into thousands of underlying venture deals across decades — the shape was brutal and consistent: roughly six percent of deals produced around sixty percent of the returns, and about half of all deals returned less than the money invested. Not half of the bad funds’ deals. Half of everyone’s deals, including the great funds’. In fact the funds with the best overall returns tended to have more strikeouts than the mediocre ones, not fewer — because they were swinging at things that could actually be outliers.

Peter Thiel makes the same point in Zero to One, and his formulation is the one worth memorizing: in a venture portfolio, the best investment tends to equal or outperform the entire rest of the fund combined. The distribution isn’t a bell curve with a fat middle. It’s a power law, and a power law has no meaningful middle. There are outliers and there is noise.

If you’ve read part one of this series, you can already see the collision coming. Everything a well-run corporation knows about risk — diversify, avoid losses, weight the downside, demand a forecastable base case — is calibrated for normal distributions, where errors cancel and prudence compounds. Point that machinery at a power-law asset class and it does exactly what it was built to do: it filters out variance. Which means it filters out the outliers, which were the only source of returns. In venture, the catastrophic error isn’t backing a failure. It’s passing on the outlier. The spreadsheet that protects the core guarantees mediocre venture returns, and it does so while producing impeccable documentation of its own prudence.

So the picking question — the subject of this post — isn’t “how do I avoid backing losers?” You will back losers; the math says roughly half your portfolio, if you’re doing it right. The question is: what actually predicts the outliers, what’s folklore, and what does a corporate investor specifically get wrong? Part seven covered how to structure a CVC so it can invest at all. This one is about the craft of the investment decision itself.

Timing: the factor nobody’s model includes

The best-known attempt to actually measure what predicts startup success, rather than theorize about it, came from someone with an unusual dataset: his own companies. Bill Gross founded Idealab in 1996 and, by his 2015 TED talk, had watched over a hundred companies born inside it plus a comparison set of about a hundred outside ones — successes and failures in both groups. He scored roughly two hundred companies on five factors — idea, team, business model, funding, and timing — and asked which factor best separated the winners from the dead.

Timing came first, explaining about 42% of the difference in his analysis — ahead of team and execution, ahead of the idea itself, with funding dead last. His examples are the memorable part. Idealab’s Z.com, an online entertainment company, died in 1999–2000 when broadband penetration was too low and video codecs too painful; YouTube arrived in 2005, after Flash and broadband had quietly solved both problems, and won with no business model at all at launch. Airbnb and Uber launched into a recession that made ordinary people willing to rent out rooms and drive strangers for money — timing that looked like a bug and was actually the feature.

Treat Gross’s percentages as directional, not gospel — it’s one investor scoring his own portfolio, not a controlled study. But the operational lesson survives any quibble with the method: the most important diligence question is “why now?” Not “is this a good idea?” — most fundable ideas have been tried before and died. The question is what changed: a cost curve crossing a threshold, a platform shift, a regulation, a behavior the last crisis normalized. A founder with a crisp, falsifiable why-now answer is telling you they understand they’re surfing a wave rather than claiming to be the ocean. A founder whose answer is “nobody’s thought of this” is telling you they haven’t read the graveyard.

This is also, incidentally, the one question where a corporate investor can be genuinely better than a financial VC — you have operating data about whether the enabling shift is real. More on that below.

Jockey or horse? What the evidence actually says about teams

Ask any VC what they invest in and you’ll hear “people.” The Gompers, Gornall, Kaplan, and Strebulaev survey of nearly nine hundred VCs — published in the Journal of Financial Economics — confirms the self-report: the team was the factor investors ranked most important, in both selecting deals and explaining outcomes, well ahead of business model or market.

But there’s a wrinkle in the academic record that the folklore skips. Kaplan, Sensoy, and Strömberg tracked companies from early business plan through IPO in a 2009 Journal of Finance paper pointedly titled “Should Investors Bet on the Jockey or the Horse?” Their finding: the business line — what the company actually sells, to whom — stayed remarkably stable from founding to public listing, while the management, including founders, frequently turned over along the way. The horse persisted; the jockeys changed. Their conclusion was that investors, at least at later stages, should weight the business more than the people, because the business is what they’ll still own in year eight.

I don’t read these two results as a contradiction so much as a correction of which team qualities matter. VCs say “team” and then, in practice, often score charisma, pedigree, and confidence — traits that predict fundraising ability much better than they predict outcomes. What the evidence and my own scar tissue point to instead is two narrower things. First, rate of learning: how much did this team’s understanding of the customer change in the last six months, and can they show you the evidence that changed it? A team that recites the same pitch it gave a year ago, in a market this uncertain, has stopped learning, and polish is not a substitute. Second, founder-market fit: does this specific team have unfair insight into this specific problem — years inside the industry, a technical capability few possess, a lived version of the customer’s pain? Charisma is evenly distributed across the winners and the corpses. Earned insight is not.

Markets: Don Valentine’s heresy, and TAM theater

Sequoia’s founder Don Valentine spent his career cheerfully committing heresy against the people-first consensus: bet on markets. His formulation was that he looked for markets so large and so hungry that they would pull the product out of the company — even a flawed team can win in a great market, and Valentine liked opportunities big enough that the management couldn’t screw them up. Andy Rachleff, who co-founded Benchmark, later sharpened the same idea into what he only half-jokingly called a law: when a great team meets a lousy market, the market wins; when a lousy team meets a great market, the market wins; only when a great team meets a great market does something special happen. Rachleff’s framing is the ancestor of the product-market-fit vocabulary everyone now uses — his point being that PMF is mostly a property of the market’s desperation, which the team either finds or doesn’t.

The corporate failure mode here has a specific shape, and it’s the pitch-deck slide everyone has seen: the $50 billion top-down TAM, sliced with an “if we capture just 1%” incantation. Call it TAM theater. Top-down market sizing tells you nothing about whether anyone desperate exists, because it’s constructed by multiplying analyst reports rather than by counting customers. The diligence that actually discriminates is bottom-up: who, by name or by segment, has this problem badly enough to pay this price, how many of them exist, how do you reach them, and what does the arithmetic of price times reachable buyers actually sum to? A bottom-up number is usually embarrassingly smaller than the deck’s TAM — and infinitely more informative, because the startup that knows its first two hundred customers by type is describing a market it has touched, while the one waving at Gartner is describing a market it has read about.

One more twist that matters enormously for corporate investors: the great venture markets usually didn’t look great at entry. They looked small, weird, or beneath notice — that’s why the entry was cheap and the incumbents absent. This is the disk-drive pattern from part one viewed from the other side of the table. Which sets up the corporate-specific trap.

Thesis, opportunism, and the discipline of the anti-portfolio

There’s a long-running argument between thesis-driven investing — develop a view about where a sector is going, then hunt for the companies that view implies — and opportunistic investing, which takes strong companies as they come. For a corporate investor I think the answer is unambiguous: you need a thesis, because a thesis is the only defense against your deal flow, which will otherwise consist of whatever bankers and accelerators decide a corporate will pay a strategic premium for. A written thesis also creates something precious: the ability to be wrong on the record, and to notice.

Which brings me to the single practice I’d import from venture into any corporate development team: the anti-portfolio. Bessemer Venture Partners publicly maintains a page of the great companies it evaluated and declined — Apple, Google, Facebook, Airbnb, PayPal, Intel — complete with the self-lacerating reasoning at the time (the Google pass famously involved a partner asking a friend how he could avoid the two Stanford students renting her garage). The page is funny, and the discipline underneath it is deadly serious. In a power-law business, your errors of omission are invisible by default — nobody runs a post-mortem on a deal that never happened — yet they’re precisely the errors that determine returns. Keeping an anti-portfolio forces the review: what did we pass on, why, and what did our reasoning systematically screen out? If your passes cluster around “too small,” “no synergy,” and “unproven market,” you haven’t been prudent. You’ve been running the corporate immune system from part one with a term sheet in its hand.

Strategic fit is a tiebreaker, not a filter

Now the corporate-specific layer, and the mistake I’ve watched more than once: the CVC whose screening memo requires every investment to map to a current business-unit priority. It sounds like discipline. It’s actually a category error — it converts the venture portfolio into an outsourced feature roadmap, restricted to what the three-horizons post would call H1 territory: things adjacent enough to today’s business that the strategy department can already name them. But if the startup fits today’s roadmap that cleanly, you usually shouldn’t invest in it — you should build it, buy it outright, or sign a commercial deal. The unique value of a minority investment is optionality: cheap exposure to futures your planning process can’t yet underwrite, exactly the logic laid out in the innovation-accounting post. Options on things you’re already certain about are worthless; you’re paying a premium for information you already have.

So invert the order of operations. Screen on venture merits — the why-now, the learning rate, the desperate market — and use strategic relevance as the tiebreaker between deals that already clear that bar, plus a source of the proprietary insight financial VCs lack. And re-read the strategic-strings warning from part seven before every term sheet: the moment your investment committee starts demanding exclusivity, rights of first refusal, or roadmap commitments as the price of the check, you’ve stopped buying optionality and started taxing the startup’s other options — which the best founders, who have alternatives, will simply decline. Adverse selection does the rest: strategic strings filter your deal flow down to the companies desperate enough to accept them.

Diligence: what you can uniquely verify, and how you’re uniquely blind

A corporate investor does hold two genuine diligence advantages, and both are underused. First, customer truth: you are the customer, or you sell to thousands of them. Where a financial VC calls five references the founder supplied, you can put the product in front of real buyers inside or alongside your own business and watch what happens — the difference between polling and a live experiment. Second, technical truth: your staff engineers, chemists, and operators can pressure-test claims in an afternoon that a generalist associate would take on faith. When the pitch says the new process cuts unit cost forty percent, someone in your plant knows which assumptions in that number are load-bearing.

But the same proximity is the bias. Your experts will evaluate the startup against your architecture, your customers, your cost structure — and will therefore be most accurate about sustaining innovations and most reliably, confidently wrong about disruptive ones. This is the incumbent’s curse in diligence form: Seagate’s marketers, remember, tested the 3.5-inch drive against exactly the wrong customers and got a rigorous, well-documented, fatal answer. When your internal expert says “our customers would never accept this,” treat it as data about your customers, not about the market. The practical countermeasure is structural, not attitudinal: have the venture case argued by someone whose mandate is the option value, circulate the internal expert’s critique alongside an explicit “what would have to be true for this critique to be the Seagate memo” section, and track both in the anti-portfolio. You will still miss outliers. The goal is to stop missing them for reasons a filing cabinet already contains.

What happens after the check

None of this picking craft matters if the venture dies of the two ordinary causes. The evidence on selection is humbling precisely because so much of the outcome is decided after the investment, by operating discipline the cap table can’t supply: whether the venture prices for the value it creates rather than the costs it incurs, and whether it manages its cash to the point where survival stops depending on anyone’s continued enthusiasm — yours included. Those are the next two posts: pricing strategy for new ventures, and runway, burn, and the default-alive question. Pick in a power-law world, then operate like the power law owes you nothing — because it doesn’t.

Further reading

• Peter Thiel, Zero to One (2014) — the “Follow the Money” chapter is the cleanest statement of power-law portfolio logic in print.
• Benedict Evans, “In praise of failure” — the Horsley Bridge data on hit rates, loss rates, and why the best funds strike out more.
• Bill Gross, “The single biggest reason why start-ups succeed” (TED, 2015) — the Idealab timing analysis; watch it for the Z.com/YouTube pairing.
• Kaplan, Sensoy & Strömberg, “Should Investors Bet on the Jockey or the Horse?” (Journal of Finance, 2009) — the horse persists; the jockeys change.
• Gompers, Gornall, Kaplan & Strebulaev, “How Do Venture Capitalists Make Decisions?” (Journal of Financial Economics, 2020) — what nearly nine hundred VCs say they do.
• Bessemer Venture Partners, “The Anti-Portfolio” — the public list of great passes, and a discipline worth stealing whole.