A secondary sale of the best position swaps marked value for cash at a discount, and the right test is the break-even exit multiple, not the DPI it adds.
Selling part of a winner in a secondary buys DPI with TVPI, and the trade is worth it only if the stake is expected to grow by less than the investors' cost of capital net of the discount. On an illustrative $100 million fund, selling 30 per cent of a $90 million position at a 20 per cent discount lifts DPI from 0.40x to 0.62x and cuts TVPI from 2.20x to 2.15x. Holding wins only if the stake is worth more than 1.259 times today's mark in four years, which a power-law outcome reaches on average but misses 70 per cent of the time.
Worked in full in Venture Capital by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
A fund in its seventh year has called its whole commitment, distributed little and carries most of its value in one company. The investors want cash back; the next fundraise needs a DPI that does not start with a zero. A late-stage investor offers to buy part of the fund's stake in the best company at a discount to the latest round price. The partners have to decide whether to sell, and how much. Throughout, the proceeds are assumed to be distributed to the investors straight away; cash kept in the fund adds nothing to DPI.
| Item | Value |
|---|---|
| Paid-in capital | 100.0 |
| Distributions to date | 40.0 |
| Net asset value | 180.0 |
| of which the winner, at the last round price | 90.0 |
| Share of the stake offered for sale | 30% |
| Discount to the mark | 20% |
| Investors' required return on cash returned | 12% |
| Years to the expected exit of the winner | 4 |
Before the sale the fund reports DPI 0.40x, RVPI 1.80x and TVPI 2.20x, and the winner is 50.0 per cent of NAV.
Mark of the stake sold = 90.0 × 30% = 27.0
Proceeds = 27.0 × (1 − 20%) = 21.6; discount given up = 5.4
DPI = (40.0 + 21.6) / 100.0 = 0.616x, up 21.6 points
TVPI = (40.0 + 21.6 + 180.0 − 27.0) / 100.0 = 2.146x, down 5.4 points
Break-even exit multiple of the mark = (1 − discount) × (1 + r)n = 0.80 × 1.124 = 1.259
Excel: =(1-Disc)*(1+r)^n for the break-even; =(Dist+Mark*Share*(1-Disc))/PaidIn for the new DPI
The DPI gain is exactly the proceeds and the TVPI loss is exactly the discount, because the fund swaps 27.0 of marked value for 21.6 of cash. Nothing else on the report moves. NAV falls to 153.0, the winner to 63.0 and its share of NAV to 41.2 per cent.
The break-even is the line that matters. If the investors can earn 12 per cent a year on cash returned, the 21.6 is worth 33.99 in four years. Holding the stake beats that only if the 27.0 of mark becomes more than 33.99 at exit, a multiple of 1.259 on today's price, or 5.9 per cent a year of growth in the company's value. A later-stage company the fund still believes in usually clears that bar on paper, which is why partners rarely want to sell.
Give the stake three outcomes in four years: 3.0x the mark with 30 per cent probability, flat at 1.0x with 40 per cent, and 0.3x with 30 per cent.
| Outcome | Probability | Stake worth | Versus selling |
|---|---|---|---|
| 3.0x the mark | 30% | 81.0 | Holding wins |
| 1.0x the mark | 40% | 27.0 | Selling wins |
| 0.3x the mark | 30% | 8.1 | Selling wins |
| Expected | 37.53 | Holding wins by 3.54 |
The expected multiple is 1.39, above the 1.259 break-even, so the arithmetic of the mean says hold. But in 70 per cent of outcomes selling leaves the investors better off. That is the power law at the level of a single position: the mean is carried by the one outcome in which the company becomes very large, and a fund that sells part of the stake is giving up a slice of exactly that outcome in exchange for certainty in the other two.
The decision is not about whether the company will do well. It is whether the fund wants a 1.39x expected multiple with a 70 per cent chance of underperforming cash, or a sure 1.259x equivalent, on this slice of its biggest position.
| Discount | Proceeds ($m) | DPI after | TVPI after | Break-even multiple |
|---|---|---|---|---|
| 10% | 24.3 | 0.643 | 2.173 | 1.416 |
| 20% | 21.6 | 0.616 | 2.146 | 1.259 |
| 30% | 18.9 | 0.589 | 2.119 | 1.101 |
| 40% | 16.2 | 0.562 | 2.092 | 0.944 |
A deeper discount lowers the bar that holding has to clear. At 40 per cent off, the stake can lose value and still beat selling: the break-even falls to 0.944. Without any discount the bar is the full 1.574 that 12 per cent compounds to in four years, which is why a sale at the round price is almost always worth taking for part of a position.
| Share sold | Proceeds ($m) | DPI after | TVPI after | Winner, % of NAV |
|---|---|---|---|---|
| 20% | 14.4 | 0.544 | 2.164 | 44.4 |
| 30% | 21.6 | 0.616 | 2.146 | 41.2 |
| 50% | 36.0 | 0.760 | 2.110 | 33.3 |
To take DPI to 1.0x from this one sale the fund would have to sell 83.3 per cent of the position, which is no longer trimming a winner but exiting it. How long a fund usually takes to reach 1x without such a sale is set out in how many years it takes a VC fund to reach 1x DPI.
The 5.4 of discount is fund profit given up. With TVPI well above the point at which carried interest switches on, every dollar of profit is shared 80 to 20 under a whole-fund waterfall with a full catch-up, so the investors bear 4.32 and the general partner 1.08. That alignment is weaker than it looks: the partners gain from a higher DPI in the next fundraise, and the investors do not share that benefit.
Reading the sale as a pure DPI improvement. The report shows DPI up 21.6 points and TVPI down only 5.4, which looks like a bargain; but DPI is not value, it is timing. The right test compares the cash with what the slice of stake is expected to become, discounted at the investors' cost of capital, and then looks at the distribution of outcomes, not only the mean. The mirror case, an investor selling its whole fund interest at a discount, is worked in does selling an LP interest at a discount raise DPI. The fund economics behind both, multiples, carry and the power law, are reproduced in the free workbooks for the book's fund.
Compute the break-even exit multiple, (1 − discount) × (1 + r)n, here 1.259, and set it against an honest distribution of outcomes for the stake. Selling a slice is sensible when the expected multiple is close to the break-even or the position dominates NAV; at a 1.39x expected multiple it is a choice to give up some mean for a lot less variance.
Yes, by exactly the proceeds over paid-in capital, while TVPI falls by the discount. On an illustrative $100 million fund, selling 30 per cent of a $90 million stake at a 20 per cent discount brings in $21.6 million: DPI rises from 0.40x to 0.616x and TVPI falls from 2.20x to 2.146x. The fund has changed the timing of its value, not added to it.
It varies with the company, the round and the market, so treat any figure as illustrative. What a discount means is computable: at 20 per cent, with a 12 per cent cost of capital over four years, holding wins only above 1.259 times the mark; at 40 per cent the break-even falls to 0.944, so the stake could lose value and still beat selling.
Enough to change the risk without giving away the outcome that carries the fund. In the illustrative case, selling 30 per cent cuts the winner from 50.0 to 41.2 per cent of NAV for a 5.4 point TVPI cost; reaching 1.0x DPI from that position alone would mean selling 83.3 per cent, which is an exit rather than a trim.
This article is one calculation from Venture Capital. The book takes the same case from first principles to the decision, chapter by chapter, and every figure it prints is a live formula in the free companion workbooks.
Get the book on Amazon →Free companion files
Also on Amazon UK · Amazon Germany · Amazon France · Amazon Canada
Reading guide: private equity and private markets → · All 1614 articles →
If this book helped, or didn’t, a few lines on Amazon are worth more than they look: they are what the next reader goes on. Write a review. The workbook stays free either way.