← All logs
Investors · 7 min read

The 20 Percent Rule: How Venture Funds Decide What to Hold Back

Emrah ·

Fred Wilson explains why a venture fund's real skill isn't picking winners once. It's knowing how much capital to save for the rounds that come after.

The Story

In August 2019, Fred Wilson, co-founder of Union Square Ventures, sat down to write about a problem most outsiders never think about when they picture venture capital. People imagine a VC as someone who writes one check into a promising company and waits. Wilson's own firm doesn't work that way, and neither do most serious early-stage investors.

USV frequently invests in the same company four to six separate times over five to seven years, moving from an early round through later stages as the company grows. Each of those checks has to come from somewhere, and a venture fund does not get to raise new money mid-stream. It has a fixed pool of capital, decided once, at the start of the fund's life. Every dollar spent on a follow-on round is a dollar that came out of a budget set years earlier, often before anyone knew which companies in the portfolio would actually turn into winners.

That forces a question most founders never hear discussed out loud: how much of the fund should sit untouched, waiting to be deployed into the companies that turn out to matter most. Wilson's answer, in his post titled simply "Reserves," is that this question deserves the same rigor as picking the company in the first place.

The Idea

Wilson ranks capital allocation and reserve management third on the list of skills that separate strong venture funds from average ones, right behind picking the right companies and doing good work on the board. It's a skill he thinks gets far less attention than it deserves, partly because it happens quietly, inside spreadsheets and models, long after the exciting part of writing a first check is over.

USV's approach is specific. Rather than setting aside a flat percentage based on gut feel, the firm runs a Monte Carlo simulation, modeling the portfolio's likely outcomes a thousand times over, to estimate how much follow-on capital it will actually need and when. The output of that process, Wilson says, is that roughly 20 percent of the fund gets held back and recycled specifically for these future rounds.

The logic underneath this is venture math in its purest form. Returns in early-stage investing follow a power law: a small number of portfolio companies generate the overwhelming majority of a fund's total return. If a fund fails to keep buying into its own winners as they raise more money, its ownership stake gets diluted with every new round, and the fund ends up owning less and less of the exact companies it most needs to own more of.

How Founders Should Read This

Founders almost never see this calculation happening, and that is exactly the problem. A founder who raised a strong seed round often assumes, reasonably but incorrectly, that the same investor will show up with a proportional check at the next stage. Sometimes that assumption holds. Sometimes it doesn't, because the fund's reserve for that specific company was already committed, reduced, or reallocated somewhere else in the portfolio, and nobody told the founder until close to the round.

The practical fix is uncomfortable but simple. A founder heading into a new round should ask existing investors a direct question, early: do you have reserve set aside for this company, and do you plan to use your pro-rata right this round. Investors will not always volunteer a clear answer, because a wavering follow-on decision is not something most VCs want to state plainly to a founder mid-relationship. But asking removes the guesswork, and gives the founder time to build a backup plan, whether that means courting a new lead or adjusting how much runway the round actually needs to cover.

There's a second, quieter lesson here. Founders often treat every investor conversation as equally weighted, when in practice each investor is running a private, ongoing calculation about how much more capital that specific company deserves relative to everything else in the fund. Understanding that this calculation exists, even without seeing the numbers inside it, changes how a founder should read investor behavior at each stage.

How Investors Should Read This

For an investor, reserve strategy is where fund discipline either holds or quietly breaks down. The temptation in a warm, ongoing relationship with a founder is to keep writing checks because saying no feels personal and damaging. Wilson's framework pushes against that instinct. Reserve capital exists to be deployed toward the companies most likely to drive fund returns, not distributed evenly to preserve every relationship in the portfolio equally.

That means an investor evaluating a follow-on decision has to separate two questions that often get blurred together: does this founder deserve continued support as a person and operator, and does this specific company represent the kind of outsized outcome that justifies pulling reserve capital away from something else in the fund. Those two answers are not always the same, and a fund that cannot tell them apart ends up spreading its reserve too thin across companies that will never return the fund, at the direct expense of the ones that could.

Running a formal model, rather than relying on instinct or a flat percentage rule, is Wilson's answer to this tension. A thousand simulated outcomes force an investor to confront, in advance, how uncertain the future actually is, and to size reserves against a distribution of scenarios instead of a single hopeful one. It's a less comfortable way to work than trusting your gut about which founders you like best, and that discomfort is precisely the point.

Where the Two Views Collide

The sharpest tension in this entire framework is informational, not just financial. A founder wants to know, as early and clearly as possible, whether their existing investor will keep backing them. An investor has strong reasons not to answer that question directly or early, because revealing a weak reserve position changes how a founder negotiates, who else they approach, and how urgently they start building a backup plan.

This creates a strange asymmetry at the exact moment founders most need clarity. The investor who has quietly decided to direct reserve elsewhere in the portfolio has every incentive to stay vague for as long as possible, since an ambiguous answer buys the fund time and keeps the founder from immediately raising elsewhere on worse terms. Meanwhile, the founder who assumes support without asking is operating on hope dressed up as a relationship.

A founder who understands this dynamic stops treating "our investor will follow on" as a fact and starts treating it as a hypothesis worth testing directly, months before the round actually needs to close. An investor who understands it recognizes that a founder pushing hard for a clear reserve answer isn't being paranoid. They're doing exactly what Wilson's own model does: replacing a comfortable assumption with an actual estimate.

The Takeaway

Reserve math is the part of venture investing that never makes it into a pitch deck or a congratulatory tweet about a new round, but it quietly decides which companies get the support they need to survive to their next milestone and which ones get left to find it elsewhere. For founders, the lesson is to ask the reserve question directly, early, and often, instead of treating investor support as a given. For investors, the lesson is that a fund's discipline shows up less in which companies it picks at the start and more in how honestly it manages the capital it has left for the companies that turn out to matter.

The next time you hear that a company has a strong lead investor behind it, the real question isn't whether that investor believes in the founder. It's whether that investor actually has the reserve left to prove it.


Source: Fred Wilson, "Reserves," AVC, 2019 — Reserves

Past editionsLondonBerlinIstanbulAmsterdam
LegalTerms & Privacy
© 2026 Prompt Network Inc.
Prompt

New York is open. AI, 16 to 17 November.

Apply now