Scenario Engine and Follow-On Strategy

Has the 40 to 50 percent reserve rule actually broken down

For most of the last fifteen years an early-stage fund held back 40 to 50 percent of its capital for follow-on, and almost nobody argued about it. That convention is now contested, and not as a single dispute with two sides: there are at least three distinct positions, each with a different view of what a reserve is actually for.

The convention, and why it is contested

For most of the last fifteen years, an early-stage fund held back something like 40 to 50 percent of its capital for follow-on investment. The reasoning was straightforward. A venture portfolio returns through a small number of outliers, and a fund that identifies one early and cannot put more money into it has converted its best insight into its smallest position. The reserve pool was the mechanism for avoiding that.

That convention is now contested, and the disagreement is not a single dispute with two sides. It is at least three distinct positions, each with a different argument and a different view of what a reserve is actually for. Reading them as schools rather than as a running argument makes it easier to work out which one describes your fund.

The reserve-light school

The position

Below a certain fund size, reserves destroy more value than they create, and the capital does more work deployed into initial positions. Hunter Walk states it at its strongest: early-stage funds of $100 million or less should hold almost no reserves for follow-on.

Four arguments carry the position. Follow-on rounds happen too quickly to give a fund reliable signal about which company is the real outlier, so the picking insight the reserve was meant to exploit is weaker than it used to be. Pro rata rights have become partly illusory on crowded cap tables where much larger funds want the allocation. Pricing discipline is broken by multistage firms deploying at return targets a small fund cannot match. And the historical data underwriting the practice describes a market that no longer exists.

The concentration figures behind the pro rata argument are the part hardest to dismiss. The Fund CFO, drawing on the PitchBook-NVCA Venture Monitor, records three firms taking in 48.1 percent of all venture capital raised in the first half of 2026, and megadeals of $100 million or more absorbing 87.5 percent of the $412.7 billion deployed. A small fund holding reserves to defend its position in a hot round is holding them against counterparties for whom its participation is a rounding error.

The probability school

The position

A reserve should be sized against the probability that the capital lands where it needs to, rather than set by convention. Where that probability is low, the reserve is a bet on a coin flip the fund cannot win often enough.

Andreas Schmidt makes the sharpest version of this, and it is uncomfortable because it is about selection rather than market structure. For a fund with a thirty-company portfolio that follows on into five companies, he puts the probability of selecting the eventual winner at random at less than 17 percent. Reserves only pay when they are concentrated decisively into the right company, and concentration is precisely what makes the odds of being right so poor. His conclusion is that funds up to roughly €25 to €30 million should deploy all their capital into initial positions instead.

The same logic runs through graduation rates, which is the other input the probability is built on. A reserve held for a round that never happens is capital that sat idle for a decade, and the share of companies that reach the next round is therefore a direct input to how much reserve is worth holding. This is the school that treats the reserve percentage as an output of the arithmetic rather than a starting assumption.

Note that this school and the reserve-light school reach similar conclusions for small funds by different routes. One argues from the market a small fund operates in; the other from the odds a small portfolio faces. They are not the same argument, and a firm persuaded by one is not automatically committed to the other.

The reserve-preserving school

The position

The same conditions that look like an argument against reserves are an argument for them. As fewer companies graduate, the ones that do become scarcer, more clearly identified, and more contested, and the fund that cannot participate gives up its position in the only companies that mattered.

The evidence here is the same evidence. Konvoy's work on the Series A crunch, drawing on Carta's fund performance data, finds that 30.6 percent of companies raising a seed round in the first quarter of 2018 reached a Series A within two years, against 15.4 percent of the companies that raised seed in the first quarter of 2022. Graduation has roughly halved industry-wide.

The probability school reads that as most reserve dollars being held for rounds that will never happen. This school reads it the other way: if only one seed company in six reaches a Series A, then each of those is more consequential, better identified by the time it gets there, and harder to buy into. Scarcity makes each follow-on decision matter more, not less.

There is also a position that declines the choice. The Fund CFO points to recycling provisions rather than passive reserve pools: redeploying early exit proceeds lets a fund reach well over 100 percent deployed without holding capital idle for years, with Walk's own firm reportedly getting past 120 percent invested in each of its first two funds this way. Recycling changes when the capital is available rather than whether it exists, which sidesteps part of the disagreement entirely.

What the schools agree on

One claim is common ground, and it is the most useful thing on this page. Every position here is scoped by fund size. The reserve-light argument confines itself to funds of $100 million or less. The probability argument confines itself to funds up to roughly €25 to €30 million. Neither claims to describe a $500 million fund.

So the question a General Partner can actually answer is not which school is right in general. It is whether their own fund's reserve is large enough to change an outcome in a round priced by someone else. That is arithmetic about one fund, and it resolves before any of the general arguments do.

Where the Colibrí Architecture model sits

The model has no position on the right reserve percentage, and it would be a mistake to read one into it. What it does is make the consequences of the choice visible in the rest of the configuration, because a reserve decision is not a self-contained one.

Reserves are declared as a share of fund size, and that share determines how much capital is left for initial cheques. Divide what remains by the target portfolio count and you have the implied initial cheque, which either matches the stage the fund intends to enter at or does not. This is one of the dimensions Portfolio Efficiency reads. A fund that raises its reserve percentage without changing anything else has quietly reduced its initial cheque; a fund that drops reserves to near zero has raised it. Neither is wrong. Both change what the fund is.

The Scenario Engine sits on the other side of the same decision. It is a companion module, not one of the three engines, and it takes up the question after the reserve policy has been set: for a specific existing portfolio company, in a specific proposed round, should this fund follow on, how much, and when. It reads the company's trajectory, the round terms, the fund's remaining reserve capacity, and the firm's existing exposure to that company across every fund it runs, and it returns a recommendation, an allocation expressed against pro rata, a timing signal, and a three-part rationale.

It is decision support and nothing more. Each scenario is a point-in-time evaluation against current inputs. The Scenario Engine does not execute commitments, does not maintain a standing recommendation between sessions, and does not optimize or allocate a reserve pool. The General Partner makes the decision. What the module offers is that the reasoning behind it is explicit and reviewable, which matters more than usual on a question the industry itself has not settled.

Sources

Take it to your own fund

Run the model on your own fund.

The platform reads the Colibrí Architecture model against your own firm and funds.

Start a trial