Module 10 · Venture Finance

Portfolio construction and the power law

Module 04 established the power law as a fact about venture returns: a handful of investments produce nearly all the gains. This module is about what a fund manager actually does with that fact. How many investments should a fund make? How much should it reserve for follow-ons? Should it spread bets widely or concentrate on high conviction? The answers are not matters of taste — they follow from the mathematics of the power law, and getting them wrong is one of the most common ways venture funds fail.

35 minute read
8 sections
4 international cases
1 worked portfolio model
6-question quiz
Section 01

Why portfolios exist at all

If a venture investor could reliably identify the one company in a hundred that becomes a fund-returner, they would invest only in that one company and skip the other ninety-nine. They cannot. No one can. The single most important fact about early-stage investing is that the winners are not identifiable in advance — not by the best investors, not by the founders themselves, not by anyone. The power law describes the outcome distribution, but it gives no method for predicting which specific company will land in the tail.

This is why portfolios exist. Since you cannot pick the one winner, you must make enough bets that you are statistically likely to capture a winner when one occurs. The portfolio is not a hedge against risk in the ordinary sense — it's a mechanism for ensuring exposure to an outcome distribution where the rare event drives everything.

The core problem of venture portfolio construction

You know that roughly 1 in 100 early-stage companies will become a fund-returner (a 50×+ outcome). You cannot identify which one in advance. Therefore you must construct a portfolio large enough and structured well enough that you are likely to hold the winner when it appears — and large enough ownership in it that the win actually returns your fund. Portfolio construction is the discipline of making those two things true simultaneously.

Everything in this module follows from that problem. The portfolio must be big enough to capture a rare winner (which argues for many investments), but each position must be big enough that the winner actually moves the fund (which argues for fewer, larger investments). These two pressures pull in opposite directions, and resolving the tension — for a specific fund size, stage, and strategy — is the essence of portfolio construction.

Section 02

How many investments — the sizing math

Why do venture funds typically make 20-40 investments rather than 5 or 100? The answer comes from the probability arithmetic of capturing a rare winner.

Suppose roughly 1 in 50 early-stage investments becomes a fund-returner — call it a 2% hit rate (the rate varies by stage and strategy; this is illustrative). If a fund makes only a handful of investments, the chance of capturing even one fund-returner is low:

Investments in portfolio Chance of zero fund-returners Chance of at least one
5 investments90.4%9.6%
10 investments81.7%18.3%
25 investments60.3%39.7%
40 investments44.6%55.4%
100 investments13.3%86.7%

(These are computed as the binomial probability of zero successes given a 2% per-investment hit rate: chance of zero = 0.98n.) The pattern is stark. A 5-investment fund has only a ~10% chance of capturing even one fund-returner — meaning a 90% chance of failure regardless of how good the individual picks are. A 40-investment fund crosses the 50% line. A 100-investment fund is almost certain to capture at least one.

So why don't funds just make 100+ investments to maximize the chance of catching a winner? Because of the other pressure: ownership. The more investments a fund makes from a fixed pool of capital, the smaller each check, and the smaller the fund's ownership in any single company — including the eventual winner. Section 04 works through this, but the intuition is: a 100-investment seed fund might own only 2-3% of each company, so even capturing a fund-returner produces a smaller return on the fund than a more concentrated fund would have gotten from the same winner.

The two opposing forces, quantified

More investments → higher probability of capturing a winner, but smaller ownership in each (so each winner returns less to the fund).
Fewer investments → larger ownership in each (so each winner returns more), but lower probability of capturing one at all.
The optimal portfolio size balances these. For most early-stage funds, the balance lands at roughly 20-40 initial investments, though seed funds run larger (more, smaller bets) and concentrated Series A funds run smaller (fewer, larger bets).

Section 03

The fund-returner test

One of the most important disciplines in venture investing follows directly from the power law: every initial investment a fund makes should be capable, on its own, of returning the entire fund. This is the "fund-returner test," and it's a more demanding filter than it first appears.

The logic: since most investments will fail or return little, the fund's overall return depends on its few winners being big enough to carry the rest. If no single investment in the portfolio could plausibly return the whole fund, then even the fund's best outcome won't generate venture-level returns. So a disciplined GP asks, before every initial investment: "If this company becomes everything we hope, could this single investment return our entire fund?" If the answer is no, the investment fails the test, regardless of how attractive it otherwise looks.

The fund-returner test, formalized
Required exit value = Fund size ÷ (Ownership % at exit)

For a $100M fund that will own 10% of a company at exit, the company must be able to exit at $100M ÷ 0.10 = $1B for that single investment to return the fund. If the company's realistic ceiling is a $200M acquisition, the investment can return at most $20M — a fifth of the fund — and therefore fails the test. The investment might still make money, but it cannot be a fund-returner, and a portfolio built entirely of such investments cannot produce venture returns.

The test explains several otherwise-puzzling venture behaviors:

  • Why VCs pass on "good" companies. A company that will clearly succeed as a $100M business but has no path to becoming a multi-billion-dollar company often gets passed on by venture investors — not because it's a bad company, but because it fails the fund-returner test. (This is the Module 01 distinction between "good business" and "venture-backable business," now expressed in fund-math terms.)
  • Why VCs push for huge outcomes. A founder content to sell for $200M and a VC who needs a $2B outcome to return their fund are structurally misaligned. The VC will push the founder to keep going, raise more, and swing for the larger outcome — because a $200M exit, however good for the founder, doesn't move the VC's fund. (This is the Module 03 founder-VC tension, now expressed in fund-math terms.)
  • Why ownership targets matter so much. The test depends on ownership at exit. A fund that owns 2% of the eventual winner needs that winner to exit at 50× the fund size to return the fund — an almost impossible bar. A fund that owns 15% needs only ~7× the fund size. This is why funds care intensely about entry ownership and about defending it through follow-on rounds.

The fund-returner test is the single most useful lens for understanding venture investment decisions. When a VC's behavior seems irrational from a founder's perspective — passing on a profitable company, pushing for a riskier bigger outcome, insisting on a certain ownership level — it usually makes sense once you apply the test.

Section 04

Reserves and follow-on strategy

A common mistake of first-time fund managers is to deploy all their capital into initial checks, leaving nothing for follow-on investments. Experienced funds reserve a substantial fraction of their capital — often 40-60% — to invest again in their best-performing portfolio companies in later rounds. The reserve strategy is one of the highest-leverage decisions in portfolio construction.

Why reserve at all

Recall the power law: most investments fail, a few succeed enormously. The problem is that you don't know which is which at the time of the initial check — but you learn over the following 1-3 years. The companies that are clearly working start to separate from those that aren't. Reserves let a fund double down on the apparent winners once the signal is clearer, concentrating more capital into the companies most likely to be fund-returners.

The reserve logic

Initial checks are made under maximum uncertainty — you're spreading bets because you can't tell winners from losers yet. Follow-on checks are made under reduced uncertainty — the winners have started to reveal themselves. By reserving capital for follow-ons, a fund effectively gets to invest more in its winners after gaining information, partially solving the "can't pick the winner in advance" problem. A fund that reserves 50% for follow-ons is saying: "We'll make our initial bets broadly, then concentrate our remaining capital into whichever bets are working."

The follow-on math

Reserves also serve a defensive purpose: maintaining ownership. Recall from Module 07 that every new round dilutes existing holders. A fund that wants to maintain its ownership percentage in a winning company must exercise its pro-rata rights (Module 05) and invest in subsequent rounds. Without reserves, the fund's ownership in its winner erodes round after round — exactly the wrong outcome, since the winner is where the fund's returns come from.

Why follow-on protects the winner's contribution A fund owns 12% of a company after seed.
Without follow-on: after Series A, B, C dilution (~20% each), ownership falls to ~6%.
With pro-rata follow-on: ownership maintained near 12% through the rounds.
At a $5B exit, that's the difference between $300M (6%) and $600M (12%) returned to the fund.

The reserve decision is genuinely hard because it requires the fund to predict, mid-life, which of its companies deserve follow-on capital — and to resist the temptation to "average down" into struggling companies (throwing good money after bad) or to over-reserve and under-deploy. The best funds are disciplined about following their winners and ruthless about not following their losers.

⚠️ The "averaging down" trap
A natural but destructive impulse: a portfolio company is struggling and raising a down round, and the fund is tempted to invest more to "protect" its initial investment or avoid admitting the loss. This is averaging down, and it usually compounds the error. Follow-on capital should flow to the companies showing the strongest signal of becoming fund-returners — not to the companies that most need rescuing. The power law is unforgiving here: a dollar of follow-on into a likely-winner can return 20×; the same dollar into a struggling company usually returns zero. Sentiment and sunk-cost reasoning are the enemies of good reserve deployment.
Section 05

Concentration vs. diversification

One of the genuine strategic debates in venture is how concentrated a portfolio should be. The probability math of Section 02 argues for many bets; the ownership math argues for few. Different respected firms have landed in very different places on this spectrum, and both extremes have produced great returns. The debate is real, not settled.

The concentration case

  • Make fewer, larger, higher-conviction bets
  • Own more of each company (10-20%+)
  • Each winner moves the fund substantially
  • Deep partner involvement with each company
  • Risk: may miss the winner entirely (fewer shots)
  • Exemplar: Benchmark (small funds, concentrated bets, deep involvement)

The diversification case

  • Make many smaller bets
  • Own less of each company (2-10%)
  • High probability of capturing a winner
  • Less involvement per company
  • Risk: winner's contribution diluted by small ownership
  • Exemplar: Y Combinator (hundreds of bets, tiny ownership each)

The two strategies are coherent in different ways. The concentration strategy bets that the firm can pick winners better than chance — that its judgment, diligence, and involvement justify making fewer, bigger bets. If the firm really does have superior judgment, concentration amplifies it. The diversification strategy bets that winners can't be reliably picked, so the best approach is broad exposure plus the ability to follow on into whoever emerges. If picking really is mostly luck, diversification captures the upside without requiring a skill that may not exist.

The empirical evidence is genuinely mixed, which is why the debate persists. Benchmark's concentrated model produced some of the best fund returns in history (the famous $6.7M investment in eBay that returned $5B; the early Uber stake). Y Combinator's hyper-diversified model also produced extraordinary returns (funding Airbnb, Stripe, Coinbase, Doordash, and dozens of other major outcomes across thousands of small bets). Both work. The question for any given fund is which model fits its actual capabilities — and the most common error is a fund believing it has winner-picking skill (justifying concentration) when it actually doesn't (and should diversify).

🇺🇸 Anchor case · Benchmark's concentration model
Benchmark has deliberately stayed small for decades — a handful of equal partners, no junior investment staff, relatively small funds, and a concentrated portfolio of high-conviction bets. The firm's 1997 investment of about $6.7M in eBay returned over $5 billion, one of the best venture returns ever recorded. Benchmark's model depends on the partners' ability to pick and deeply support a small number of companies; it keeps funds small specifically so that carry (not management fees) drives the partners' economics and so that each investment can meaningfully move the fund. The model is the clearest counterpoint to the diversified accelerator approach — and the fact that both have produced top-tier returns is exactly why the concentration-vs-diversification debate has no settled answer.
Section 06

Loss ratios and ownership targets

Two numbers drive most portfolio-construction decisions: the expected loss ratio (what fraction of investments return less than the capital invested) and the target ownership (what percentage of each company the fund aims to hold). These two numbers, combined with fund size, largely determine how a fund must be constructed.

The loss ratio

The loss ratio is the fraction of a fund's investments that lose money or return roughly nothing. For early-stage venture it's high — often 50-70% of investments return less than 1×. The loss ratio directly determines how much the winners must return to compensate. If 60% of investments return zero and 30% return roughly capital, then the remaining 10% must generate the fund's entire return — they have to be enormous to carry the dead weight of the rest.

The ownership target

The ownership target follows from the fund-returner test (Section 03). A fund needs enough ownership in its winners that capturing one returns the fund. Working backward: for a winner to return a fund at a realistic exit valuation, the fund typically needs to own somewhere in the range of 10-20% of the company by exit (for early-stage funds). This ownership target then drives the check size, which combined with fund size drives the number of investments.

How the numbers chain together

Fund size + ownership target → check size (initial + reserves). Fund size ÷ check size → number of investments. Number of investments + hit rate → probability of capturing a winner. The whole construction is a chain: you can't independently choose fund size, ownership target, and number of investments — fixing any two determines the third. Portfolio construction is largely the discipline of making this chain internally consistent.

This chaining explains why fund size is such a consequential decision. A fund that raises too much capital relative to its strategy is forced into either too many investments (diluting the winner's contribution) or too-large checks (chasing fewer, more expensive deals). A fund that raises too little can't maintain ownership through follow-on rounds. The "right" fund size is the one whose construction chain is internally consistent for the firm's stage, strategy, and capabilities — which is why experienced LPs are skeptical of funds that grow their size dramatically between vintages without a corresponding change in strategy.

Section 07

How construction varies by fund size and stage

Portfolio construction looks very different across the venture landscape. A $30M seed fund, a $500M multi-stage fund, and a 500-company-per-year accelerator are all "venture" but construct their portfolios according to completely different math.

Fund type Investments Ownership target Reserve strategy
Micro seed fund ($30M) 30-50 5-10% Light reserves; rely on initial checks
Classic Series A fund ($200M) 20-30 15-20% Heavy reserves (50%+) for follow-on
Multi-stage fund ($1B+) 30-50 varies by stage Follow winners across stages aggressively
Accelerator (YC-style) 100s-1000s ~7% Minimal initial; selective follow-on funds
Growth fund ($2B+) 15-25 5-15% Large checks, lower loss ratio, fewer bets

A few patterns worth noticing:

  • Seed funds run more, smaller bets because winners can't be identified at all at seed, so broad exposure plus follow-on is the only viable strategy. They accept lower ownership and rely on capturing winners through volume.
  • Series A funds concentrate more because by Series A, the winners have started to reveal themselves (the company has traction). Higher conviction justifies fewer, larger, higher-ownership bets, with heavy reserves to defend ownership through later rounds.
  • Growth funds run lower loss ratios because by the growth stage, companies have proven business models — fewer fail outright. The bets are larger and the multiples lower (2-5× rather than 50×), so the construction looks more like late-stage private equity than early-stage venture.
  • Accelerators are an extreme of diversification — hundreds or thousands of tiny bets at ~7% ownership, relying entirely on volume to capture the rare winner, with separate later-stage funds to follow on into the breakouts.

International variation in construction

Portfolio construction norms vary across ecosystems. European funds have historically been more concentrated (fewer bets, higher conviction), partly reflecting smaller fund sizes and a more conservative LP base. Indian and Southeast Asian funds often construct around a smaller number of category leaders per sector, reflecting markets where a few winners dominate each category. Chinese funds in their growth era ran unusually concentrated and large, fueled by abundant local capital and corporate-VC participation. The construction math is universal, but the inputs — fund size, hit rates, exit-value ceilings — differ enough by market that the resulting portfolios look quite different.

🇮🇳 Anchor case · Peak XV's multi-stage construction
Peak XV Partners (the former Sequoia India, which became independent in 2023) constructs across a wide range of stages, from seed through growth, within the Indian and Southeast Asian markets. Its portfolio construction reflects the structure of those markets: a relatively concentrated set of category-leader bets per sector, with aggressive follow-on into the companies that emerge as winners, and a willingness to write much larger checks at growth stage into the proven leaders. The approach reflects a market where category winners tend to take a dominant share (Flipkart in e-commerce, Zomato in food delivery), making the identification of the eventual leader — and concentration into it once identified — more valuable than broad diversification across many similar companies.
Section 08

A worked example — building a portfolio model

To bring the pieces together, construct a portfolio model for a hypothetical $100M seed fund from scratch. This is the kind of model a GP builds before raising a fund, to show LPs how the fund is expected to behave.

Step 1 — Set the basic parameters

Fund parameters Fund size: $100M
Management fees over life (~16%): −$16M
Investable capital: $84M
Reserve ratio: 50% (half for initial checks, half for follow-on)
Initial-check capital: $42M  |  Follow-on capital: $42M

Step 2 — Decide check size and number of investments

Initial portfolio Target initial check: $1.5M for ~10% ownership (implies ~$15M post-money entry)
Number of initial investments: $42M ÷ $1.5M = 28 investments
Follow-on: $42M reserved to defend ownership in the ~5-8 companies that show winner signal

Step 3 — Apply the power-law outcome distribution

Project the 28 investments through a realistic power-law distribution. Assume the fund follows on into its winners, maintaining roughly 10% ownership in the best outcomes:

Outcome Companies Avg exit value Fund's ~10% stake Returns
Total loss15$0$0
Acqui-hire / small7$15M10%$10.5M
Solid exit4$150M10%$60M
Strong exit1$600M10%$60M
Fund-returner1$2B10%$200M
Total28$330.5M

Step 4 — Compute fund return

Fund-level return Total returns: $330.5M
Fund size: $100M
Gross MOIC: $330.5M ÷ $100M = 3.3×
After fees and carry, LP net: roughly 2.5× — a strong seed-fund outcome

What the model shows

  • The single fund-returner ($2B exit) contributed $200M — more than the other 27 investments combined. This is the power law from Module 01 expressed in a constructed portfolio.
  • 15 of 28 companies (54%) returned zero. The loss ratio is high and that's fine — it's structurally expected, and the model still produces a 3.3× gross return.
  • The 10% ownership was essential. Had the fund owned only 5% (because it made twice as many investments, or didn't reserve for follow-on), the fund-returner would have contributed $100M instead of $200M, cutting the total return roughly to 2.3× gross — a meaningfully worse fund.
  • The model lives or dies on the fund-returner. Remove the $2B company and the fund returns $130.5M ($1.3× gross) — a poor outcome that loses money for LPs after fees. The entire construction is a machine for ensuring the fund holds enough ownership in whichever company becomes the fund-returner, while making enough bets to be likely to hold one at all.
The construction in one sentence

Portfolio construction is the discipline of making enough bets to be likely to capture a power-law winner (Section 02), while holding enough ownership in each (Section 06) that the winner actually returns the fund (Section 03), and reserving enough capital (Section 04) to defend that ownership through the winner's later rounds. Every number in the model — fund size, check size, investment count, reserve ratio, ownership target — is chained to every other, and the whole construction succeeds or fails on whether it captures and holds a fund-returner.

Next module

Module 11 · Exit Strategies

How venture investors actually realize returns. IPO mechanics and the thin IPO market since 2021. M&A — strategic vs financial buyers. The secondary-market explosion (Forge, EquityZen, Carta). Direct listings and the fall of SPACs. The growing gap between "exit" and "liquidity," and why founders increasingly want one without the other.

Self-examination

Test your understanding

Six questions on portfolio construction, the fund-returner test, reserve strategy, and the concentration-vs-diversification debate.

Module 10 · Self-examination