Data sourced from Dakota Private Markets, the private fund performance platform powered by Dakota. Learn More | Request Access
Every quartile table starts as a pile of individual fund records. Dakota Private Markets holds 159,000+ performance records across 18,000+ private funds, and before any one of them can sit inside a benchmark, someone has to decide where it came from, what vintage the fund belongs to, what strategy it runs, how its return is calculated, and how it gets combined with everyone else's.
Each of those decisions is defensible on its own. Each one also moves the final number. Two providers can start from the same funds, make different reasonable choices at every step, and publish different top-quartile thresholds for the same vintage.
Most benchmark discussion focuses on the output. This article walks through the build itself: six steps from raw record to published quartile, and the point at each step where a benchmark can quietly go wrong.
Benchmark providers collect fund performance through a few channels: disclosures from allocators such as public pensions, submissions from fund managers, and data from consultants or administrators. Each channel brings its own tilt.
Where it goes wrong: Allocator-disclosed data leans toward managers large and established enough to win public pension mandates. Manager-submitted data leans toward funds the manager is happy to show. A benchmark built mostly from one channel inherits that channel's view of the market, and when two sources report different numbers for the same fund, someone has to decide which one wins.
What to ask: Where did the records come from, and how are conflicts between sources reconciled? For how Dakota handles this, see How Dakota Verifies Performance Data Before It Ever Reaches a Client.
Vintage is the first filter in nearly every peer group, and there is no single definition of it. Depending on the provider, vintage can mean the year of first close, first capital call, first investment, or legal formation.
Where it goes wrong: A fund that holds its first close in November 2019 and calls its first capital in February 2020 can be a 2019 vintage in one dataset and a 2020 vintage in another. Those two years sit on opposite sides of the COVID drawdown, with very different entry valuations. The fund's performance does not change. Its peer group does, and so does its quartile.
What to ask: Which event defines vintage in this benchmark, and does it match how the fund defines its own vintage?
Every fund gets a label: buyout, growth, direct lending, value-add. That label usually starts with how the manager describes the fund at launch.
Where it goes wrong: Labels are set once, but portfolios move. A growth fund that drifts into minority buyouts, or a direct lending vehicle carrying a meaningful mezzanine sleeve, keeps its original tag. Multi-strategy funds get forced into a single bucket. Size bands have no shared cutoffs across providers either, so "middle market buyout" in one dataset is not the same set of funds in another.
What to ask: Is the classification based on what the fund says it does or what it actually holds? Peer Group Construction: Why Your Benchmark Might Be Comparing Apples to Oranges covers the filters that come after the label.
Once a fund is placed, its Net IRR, TVPI, DPI, and RVPI have to sit on a common basis: net of fees and carry, the same definition of paid-in capital, and the same as-of date. The definitional gaps are covered in Why Lack of Standardized Reporting Makes Private Fund Benchmarks So Difficult. The build-specific problem is the as-of date.
Where it goes wrong: A benchmark labeled "as of June 30" often mixes funds that reported at June 30 with funds whose latest available mark is March 31 or earlier, carried forward. In a quarter when markets move sharply, those stale marks pull the cohort toward where the market was, not where it is.
What to ask: What share of the cohort actually reported at the stated as-of date?
This is the step most readers never see, and it can move the number more than any other. Once the cohort is set, there are several ways to turn it into "the benchmark," and providers do not all use the same one.
|
Method |
How it works |
What it reflects |
|
Median and quartiles |
Rank funds and take the breakpoints; every fund counts equally |
The typical manager's result |
|
Pooled (capital-weighted) |
Combine every fund's cash flows as if they were one fund |
Where the dollars went; the largest funds dominate |
|
Simple average |
Equal-weighted mean of fund returns |
Easily pulled by outliers, especially in venture |
Where it goes wrong: A pooled return answers "how did capital in this strategy perform," and a handful of megafunds can drive most of it. A median answers "how did a typical manager perform." A $300M fund measured against a pooled benchmark is largely being measured against the few funds that hold most of the capital. In small cohorts, the method used to calculate the quartile breakpoint itself can also shift the line, a point covered in Best Practices for Creating Private Fund Performance Benchmarks.
What to ask: Is this a median or a pooled figure? If pooled, which funds carry the most weight?
Many benchmarks do not stop at private peers. Allocators also want to know whether a fund beat what the same dollars would have earned in public markets.
Where it goes wrong: The common shortcut sets a fund's IRR next to an index's annualized return. Those are not the same kind of number. IRR is money-weighted and depends on when capital went in and came out. An index return is time-weighted and assumes the money was invested the whole time.
Public Market Equivalent (PME) methods close that gap by running the fund's actual cash flows through the index: each capital call buys the index, each distribution sells it. A Kaplan-Schoar PME above 1.0 means the fund beat the index on a like-for-like basis. The index choice matters too. A lower-middle-market buyout fund measured against the S&P 500 is being compared to a very different risk profile than one measured against a small-cap index.
What to ask: Is the public comparison a PME or a straight IRR-versus-index comparison, and which index was used?
Every benchmark claim can be tested against the same six questions.
|
Build step |
Question to ask |
|
Sourcing |
Where did the records come from, and how are conflicts resolved? |
|
Vintage |
Which event defines vintage: first close, first call, or first investment? |
|
Strategy |
Is the label based on what the fund says or what it holds? |
|
Calculation |
What share of the cohort reported at the stated as-of date? |
|
Aggregation |
Median or pooled, and how many funds are in the cohort? |
|
Public comparison |
PME or straight IRR versus an index, and which index? |
When you cite a benchmark in a deck or a DDQ, put the build next to it: the source, the vintage definition, the cohort size, median or pooled, and the as-of date. One footnote line is enough.
Consultants and allocators running diligence will ask these questions anyway. A manager who answers them up front moves the conversation from "is this number right" to "what does this number say," which is the conversation worth having.
Dakota Private Markets covers 18,000+ private funds and 159,000+ performance records across seven asset classes, with every record reviewed by Dakota's research team before publication. Filter by vintage year, strategy, geography, fund size, and portfolio company sector, see the fund count behind every quartile, and export the full table for the diligence file.
Request access to Dakota Private Markets.