Five routes, and the choice among them changes reported performance even when the total proceeds do not.
What are the main private equity exit routes?
A trade sale to a strategic buyer typically pays the highest headline price, because a strategic can fund the price out of synergies a financial buyer cannot access, and it usually settles in full at closing, which is the cleanest possible outcome for distributions and for the waterfall.
A sponsor-to-sponsor sale transfers the asset to another private-equity fund. The price reflects what a leveraged buyer can pay rather than what a strategic can, and the same asset then appears in a second fund's portfolio with a second set of fees on it, which is a fact worth holding when reviewing a strategy's aggregate cost to investors.
An initial public offering rarely produces a clean exit. The sponsor sells a slice at listing and remains subject to a lock-up, typically six months, after which the remaining stake is sold down over quarters into the market. Reported performance therefore records a partial realisation and a residual mark, and the ultimate outcome depends on a price the sponsor does not control.
A continuation vehicle moves one asset, or several, out of an ending fund and into a new vehicle backed by new investors, with existing investors offered a choice between cash and a roll. Distributions are recorded, the internal rate of return is crystallised on the assets sold, and the general partner keeps managing the asset with a fresh fee and carry clock. The structure is legitimate and the conflict is obvious, which is why the price is normally set by a third-party fairness opinion and by what new money will actually pay.
A general-partner-led secondary on a portfolio does the same thing at portfolio scale, and the same conflict and the same safeguards apply.
Advertisement
Worked example: a higher multiple with a lower IRR
The mechanical point that ties the exit choice to the public market equivalent benchmarking is worth working rather than asserting. Take a position that could exit in year five for 2.50x, or partially exit for 0.75x in year five with the balance sold for 2.10x in year eight. The second route returns a higher multiple, 2.85x against 2.50x, and a lower internal rate of return, 15.95% against 20.09%, because the extra 0.35x took three additional years to arrive.
Investors compensated on multiples and investors compensated on rate of return will disagree about that decision, sincerely, and neither will be making an arithmetic mistake.
{
"skill": "Assembling the sourced inputs behind a private-company discount rate and a peer multiple",
"use": "Use a research-enabled tool to locate and date the published inputs a build-up rate needs (the current long-bond yield, the current published equity risk premium and size premium, the peer group's levered betas and R-squareds), so a model starts from something current rather than from a training-data memory.",
"tools": ["Research or browsing-enabled assistants", "A market-data terminal where one is available", "The index or study provider's own published pages"],
"prompt": "For each of the following, give me the current published figure, the exact primary source, and the as-of date: (1) the 20-year US Treasury yield and the 10-year Indian government security yield; (2) a published equity risk premium for the US and for India, naming the study and its methodology; (3) a published size premium for the smallest size decile, naming the study; (4) for these named listed peers, the levered beta and the regression R-squared, naming the index, the period and the return interval used. Do not give me a company-specific risk premium and do not give me a discount for lack of marketability.",
"verify": "Every figure must be traced to the provider's own page and date-stamped in the model, because all four move and three of them are revised annually. The two exclusions in the prompt are the point of the exercise: a company-specific premium and a marketability discount are judgments built from a written fact pattern, and an assistant will supply a confident percentage for either with no basis at all, which is the corpus's 'plausible, not true' failure applied to a number that decides what a family receives. Cross-check any beta by recomputing the total beta from the stated R-squared, and reject a peer set where the regression statistics are not disclosed alongside the beta.",
"diy": "You must be able to build the rate by hand from its five layers, state which layer each risk belongs in and why it may appear only once, compute total beta from a beta and a correlation, and say which of the three rates the buyer actually in front of you deserves. None of that is admissible from a chat window.",
"market": "IN"
}