Most people trying to find venture capital fund performance data hit the same wall: everything is either paywalled, six months stale, or aggregated so broadly it's useless for an actual decision.
Having been on both sides of the table — raising capital and deploying it across 65+ investments — I've spent years building a picture of what top performance actually looks like, where the data lives, and how to use it without getting misled by cherry-picked numbers. Here's the honest breakdown.
The Venture Capital Fund Performance Data Landscape
There is no Bloomberg terminal for VC. The data is fragmented by design — GPs want benchmark opacity, and LPs often prefer it too. Here's what actually exists:
Gold standard for LP benchmarking; data shared under NDA by GPs
Caveat: Lags 6–12 months; only top-tier fund coverage
Broadest coverage; includes fund-of-funds, secondary data
Caveat: Quality varies; relies heavily on public disclosures
Best for deal-level data; fund performance inferred from markups
Caveat: IRR/TVPI estimates are indirect; not LP-reported
Real cap table data from 40K+ companies; ground-truth valuations
Caveat: Skews toward smaller/newer funds; no legacy vintage depth
Fund-level IRR, TVPI, DPI from CalPERS, UTIMCO, CDPQ, L&I
Caveat: Not standardized; requires manual aggregation
Standardized reporting frameworks; useful for LP due diligence
Caveat: Framework only — doesn't provide actual benchmark data
What the Actual Benchmarks Show
Before you go hunting for data, it helps to know what you're looking for. These are the numbers that matter, drawn from Cambridge Associates and public pension disclosures:
Top-quartile net IRR (year 10)
25–35%
2015–2019 vintages per CA data
Median net IRR (all vintages)
10–13%
Barely clears public market equivalent
Top-quartile TVPI (year 10)
3.0x–5.0x
Best vintage years: 2015–2018
Median TVPI (year 10)
1.5x–1.8x
Majority of funds cluster here
Top-quartile DPI (year 10)
1.5x–2.5x
Cash-on-cash returned to LPs
Median DPI (year 10)
0.7x–1.1x
Most funds still largely unrealized at 10 years
The critical insight: vintage year matters more than manager brand. A 2021 fund at 1.2x TVPI in 2026 is not underperforming. A 2016 fund at 1.2x TVPI in 2026 is in serious trouble. Always normalize against the vintage year cohort, not against an abstract benchmark.
The Free Route: Public Pension Fund Disclosures
The most underutilized source of VC performance data is hiding in plain sight. Several U.S. public pension funds are legally required to disclose fund-level performance to the public. Here's where to look:
CalPERS
calpers.ca.gov
California Public Employees — annual fund-level disclosures, includes vintage, IRR, TVPI, DPI by manager
CalSTRS
calstrs.com
California State Teachers — similar transparency; often includes Sequoia, Andreessen, Kleiner Perkins vintage data
UTIMCO
utimco.org
University of Texas Investment Management — one of the most detailed public disclosures; updated quarterly
Washington L&I
lni.wa.gov
Washington State Labor & Industries — fund-level IRR and cash flows going back to the 1990s
CDPQ (Canada)
cdpq.com
Caisse de dépôt — annual report includes private equity and VC allocation breakdown and returns
What to Actually Do With the Data
Data without context is just noise. Here's the framework I use when evaluating a fund against benchmark data:
1. Anchor to vintage year
Pull the Cambridge Associates or CalPERS data for the same vintage year. A fund is only as strong as its cohort performance says it is. Top-quartile means top 25% of the same year, not top 25% ever.
2. Weight DPI over TVPI in older funds
After year 7, prioritize DPI (cash returned) over TVPI (unrealized + returned). Marks can lie. Distributions do not. A fund showing 2.0x TVPI with 0.3x DPI in year 9 is mostly story at that point.
3. Disaggregate the J-curve
Early vintages look bad on IRR because capital is deployed before returns materialize. A -5% IRR in year 3 from a good fund is normal. The same IRR in year 8 is a red flag. Context of fund age is everything.
4. Compare across managers, not just funds
A manager's Fund I might be top-quartile while Fund III lags. Benchmark each fund in isolation — do not assume performance carries over. Sequencing matters, and manager drift is real as check sizes grow.
The Limits of Benchmarking Data
I want to be honest about what this data can't tell you. Benchmarks are useful for eliminating bad managers. They are nearly useless for identifying the top 5% of performers in advance.
What benchmarks are good for
- ✓ Eliminating bottom-quartile managers from consideration
- ✓ Understanding whether a fund's claimed returns are credible
- ✓ Setting LP expectation on J-curve and cash timing
- ✓ Evaluating fee structures relative to peer funds
What benchmarks cannot do
- ✕ Predict which fund will return 10x in a new vintage
- ✕ Account for GP succession or strategy drift
- ✕ Capture qualitative edge: network, access, follow-on discipline
- ✕ Reflect the true power-law dispersion within a portfolio
The most important metric is one no benchmark captures: access. Top-quartile VC returns are concentrated in roughly 20 firms that see the best deals first. If you can't get into those funds, the benchmark discussion is almost academic. Use the data to filter — then use your network to get into the right rooms.
The data is out there. Most people just don't know where to look.
Use public pension disclosures, Carta, and purpose-built tools to benchmark VC performance without a six-figure data subscription.
Explore VC fund performance benchmarks on the VC Performance Dashboard and Benchmarking Tool at Value Add VC. Originally published in the Trace Cohen newsletter.