How much should a fund actually reserve

What funds actually do

A survey of more than sixty emerging fund managers worldwide, conducted by Blue Future Partners and published through OpenVC, found a median fund size of $56 million, a median initial cheque of roughly $850,000, and 55 percent of committed capital held for follow-on. Sapphire, looking across its own portfolio of fund managers, found most of them clustered around a one-to-one split between initial and reserve capital. Fred Wilson has written plainly that his firm reserves about half of each fund and raises the next one after deploying the other half.

So the centre of gravity is around half, and it has been for a long time. That is a real answer to the question as asked, and it is also the least useful part of the answer.

The number depends on fund size, and not gently

The clearest framing of this comes from the LP side. Alex Edelson, who allocates to emerging managers, breaks the answer into three bands rather than offering one: at $20 million and under, a preference for limited reserves or none at all; between $20 million and $40 million, a grey area where some reserve makes sense but well short of half; at $50 million and above, room to justify a ratio approaching one-to-one.

The reasoning behind the split is worth more than the bands. A reserve is only worth holding if deploying it would change an outcome. In a round priced and led by someone considerably larger, a small fund's pro-rata is often not large enough to change its own ownership meaningfully, let alone the company's trajectory. Below a certain fund size the capital does more work as another initial position, where it buys ownership at the lowest price the fund will ever see. This is the crux of the disagreement about whether the convention still holds, which has its own page in this library.

The ratio is not what determines the outcome

The most useful piece of work on this question does not answer it directly. Laura Thompson at Sapphire modelled a single fund with a single reserve strategy and varied only where the reserve capital landed. Her base case returns 5.5x gross, which becomes 4.0x net after fees and carry. Deploy the same reserve concentrated into the eventual winners and the net figure rises to 5.0x. Spread it evenly across the portfolio instead and it falls to 3.3x. Deploy it into companies that turn out not to be the winners and it falls to 2.5x.

Same fund, same reserve percentage, three outcomes spanning a range wider than most managers' entire performance dispersion. The reserve ratio was held constant throughout. What varied was selection.

This reframes the original question rather than answering it. How much to reserve matters considerably less than whether the firm can tell, at the moment it has to decide, which companies deserve the money. A firm confident in that judgment is right to hold reserves and concentrate them. A firm honest about not having that confidence yet is not being timid by holding less; it is declining to make a bet on a capability it has not demonstrated.

A calmer way to arrive at the number

Rather than choosing a reserve percentage and fitting the fund around it, the number falls out of four decisions a firm has to make anyway. The stage it enters at sets the cheque a meaningful position requires. The ownership it targets sets how large that cheque has to be. The portfolio count sets how many of them there are. Multiply through and the capital committed to initial positions is determined; whatever remains is the reserve.

Arrived at this way the reserve is a consequence rather than a preference, and it is defensible in an LP conversation in a way that a round number is not. That reframing has its own page as well, because it is the single change that most improves the quality of the answer.

What the model does with the number

The Colibrí Architecture model takes no position on the right reserve percentage and does not recommend one. What it does is read the consequence. The reserve share determines how much capital remains for initial cheques, and dividing that by the target portfolio count gives the initial cheque the fund has implicitly committed to. Whether that cheque is consistent with the stage the fund says it invests at is one of the dimensions Portfolio Efficiency evaluates.

So the model will not tell a General Partner that 55 percent is too high or 20 percent too low. It will show whether the number chosen leaves the fund able to write the cheques its own strategy requires, which is the part of the question that has a defensible answer.

Sources

• An LP take on VC portfolio constructionRodrigo Ferreira, Blue Future Partners, published on OpenVCSurvey of more than sixty emerging fund managers worldwide, with the reserve share, ticket sizes, and portfolio counts they reported.
• Dirty Secret: Venture Reserves are Not Always a Good ThingLaura Thompson, Sapphire, May 2022The scenario work showing the same reserve producing three very different net outcomes depending only on where it lands.
• LPs Seeking Alpha: Decoding the Myths and Mastery of Fund Structure, Size, Reserves, Access, and SelectionAlex Edelson on The Full Ratchet, episode 430An LP's answer keyed to fund size rather than to a single industry ratio.
• ReservesFred Wilson, AVC, January 2017The long-standing case for reserving about half a fund, from a firm that does it.