The Power Law Has Two Honest Answers

A pre-seed fund deck crossed my desk recently. Hundreds of positions, screening automated, diligence measured in minutes, the power law invoked on every page, and — rarer than it should be — honest marks. The fund was proud of its volume. Most readers would file it under spray-and-pray and move on.

That would be a mistake. The deck was coherent. That’s what made it worth taking seriously, because its coherence comes from accepting something most of the industry is still paid to deny.

Start from the two facts about venture returns that nobody disputes. Returns concentrate: a small number of positions produce most of the outcome, in nearly every dataset, in nearly every vintage. And the concentration is not reliably predictable at entry: the companies that produce the tail rarely look categorically different, at the first check, from the companies that produce the zeros. Every venture strategy is an answer to those two facts, whether or not the manager can state it that way.

There are two honest answers a manager can build.

The first is coverage. If you cannot predict which position produces the tail, hold so much of the distribution that the tail has trouble avoiding you. This is the index answer, and the deck on my desk was its pure expression. Everything that looks like laziness in a volume fund is, inside this answer, discipline: diligence measured in minutes is coherent when no single decision matters; refusing follow-on reserves is coherent because the model deliberately builds none of the information or governance apparatus that would make a second allocation decision better informed than the first; hundreds of positions are not a failure of conviction but the entire point. The indexer would rather be the house than the gambler — not betting hands, buying the field. Whether it actually gets the field is a question we’ll come back to.

The second is intervention. If prediction is insufficient, buy enough ownership, information, and governance to move the probabilities after entry. Take a position that matters, take the board seat, put operators inside the company, and spend the holding period working on the outcome instead of admiring the forecast. Allocators already underwrite this answer — three stages later. Buyout firms advertise their selectivity; Vista says it declines about 95% of the deals it assesses. But nobody asks selection to carry the entire return thesis there. Selection and price discipline are paired with contractual authority to change the asset after entry, and operational value creation is settled doctrine. The novelty in venture isn’t the doctrine. It’s applying it earlier, where a minority position and a board seat buy agency rather than control, the entry cost is lower, and the potential re-rating larger.

Both answers are honest because neither asks selection to carry the thesis alone. One responds to unpredictability by owning the distribution, the other by going to work on the piece of it that it owns. The strategies are opposite; the underlying epistemology is identical.

Now look at what most of the industry sells: a selective minority strategy that can demonstrate neither privileged access nor an intervention machine. The typical shape is twenty to forty minority positions, rarely leading, with fees priced on the third claim — that its partners can see, at entry, which companies carry the tail. Its return theory needs the tail, and it has done nothing to make the tail likely: not the breadth that gives an index its claim on the outliers, and not the demonstrated funnel advantage that would tilt the sample in its favor. It is running a smaller index without the coverage that makes indexing work, and calling the difference judgment.

Here the evidence deserves to be handled carefully, because it’s more interesting than the usual “nobody can pick” dismissal. Venture performance persistence is real — the research finds top venture firms repeat in a way buyout firms don’t, even using information an LP has at fundraising time. But when Nanda, Samila, and Sorenson decomposed the persistence, its engine wasn’t selection acumen. Initial success — including lucky initial success — buys reputation; reputation buys access; founders and syndicates route the best opportunities to the firms that already won. Roughly half of the measured persistence dissolves once you control for where, when, and in what sectors a firm invests, and the effect of early wins carries forward across dozens of subsequent investments. Meanwhile, when 885 institutional VCs were surveyed about what drives their returns, they ranked deal selection first. The industry’s self-image and its measured mechanism disagree. The returns follow the mechanism.

Access, then, deserves its own honest accounting. It is real, it persists, and it compounds. But it is not a third portfolio architecture. It is the condition that determines which distribution a manager gets to sample. It can be built — through prior results, operating reputation, domain authority, network position — but it is earned in evidence, and it cannot be asserted into existence in a data room. Once access is separated from portfolio construction, the menu of things a manager can actually build returns to two: own enough of the distribution available to you, or change the outcomes of what you own.

Watch what the largest franchise in venture actually did with its access, because the behavior is instructive. It registered as an investment adviser in 2019, walking away from the venture exemption and the limits on what it may hold. It now runs funds at every stage, wraps a services platform around the portfolio, and operates a registered evergreen vehicle built for private wealth. Whatever the firm believes about its selection, its structure no longer needs any single selection to be right — at any stage, in any sector. That is what breadth buys. The endgame of the best deal-flow machine in the industry turned out not to be sharper picking. It turned out to be being everywhere.

One more asymmetry, because the two buildable answers are not equal. Ask what each costs to copy. Coverage is assembled from inputs the market sells: capital, events, activity rankings, commodity screening models. Small checks buy broad, low-friction access to the uncontested bulk of the market — but not an unbiased sample. Allocation is rationed exactly where demand concentrates, and demand is a noisy signal: far too noisy to pick by, which is why selection stays unreliable, yet more than enough to bias a sample, because the part of the distribution the fund misses is not random. This is where the house aspiration meets its problem. A coverage fund with a biased funnel is not the house. It’s a gambler with a spreadsheet.

Even executed cleanly, coverage’s inputs are public, and the last cycle’s data-driven sourcing wave showed how quickly public-signal edges converge into the new baseline. Intervention cannot be made credible by a fundraise alone. Its inputs are people who have built revenue engines, playbooks proven across companies, and the information position that only years of governance produce — and it carries costs the index never pays: it doesn’t scale, it concentrates risk, and when an intervention fails, the failure has the intervener’s name on it. Of the two buildable answers, one is a commodity in waiting. The other compounds.

None of this makes the middle fund’s partners fools; many are excellent, and excellence isn’t the variable. The variable is structural. The selective minority portfolio bears the concentration of selection without the mechanism that makes concentration rational. It is a bet that the fee structure outlives the arithmetic.

So the allocator’s question for any venture manager reduces to one line: which engine am I buying? If the claim is access, the diligence is the funnel — the contested rounds won, the inbound quality, the syndicate position, earned and shown rather than asserted. If the claim is coverage, the diligence is the sample — its size and its bias, because a biased field isn’t a field. If the claim is intervention, the diligence is the record — companies whose trajectories demonstrably changed, not boards demonstrably attended. And if the claim is selection, unaccompanied by any engine that would explain why this firm sees or wins what others don’t, the persistence research has already priced it.