Key Takeaways

  • The venture universe expanded from 500 to more than 4,000 firms, while Adams Street Partners data reveals that just 7% of portfolio companies generate all market returns.
  • Re-up underwriting starts the day after signing a fund commitment, running as a three-year continuous diligence loop rather than an episodic sprint.
  • Institutional LPs face a four to five-year lag before acquiring statistical confidence in the quality of any individual venture investment decision.
  • Adams Street runs senior team offsites every January and June to dissect its anti-portfolio, auditing what the firm passed on three to five years prior.

The Three-Year Diligence Window

Most limited partners evaluate venture funds in compressed bursts. A general partner announces a fundraise, opens a data room, and gives investors 60 to 90 days to deliver an allocation check. Brijesh Jeevarathnam views that approach as broken. In an asset class where more than 4,000 firms now compete for capital, compressed reviews only guarantee surface-level checks.

Adams Street treats the re-up process as a continuous three-year audit. “Our reup on a venture fund begins the day after we commit to the current one,” (play from 15:34) Jeevarathnam said. “So in other words, it's a three-year process before they raise the next fund. We're in continuous DD on the next fund.” (play from 15:39)

This continuous scrutiny exists because returns in early-stage tech are slow to show themselves. Paper markups during seed and Series A rounds obscure whether a fund manager owns durable winners or temporary momentum. “In the venture LP part of the business, it takes I would say realistically four to 5 years before we have high confidence on the quality of the first decision,” (play from 16:47) Jeevarathnam noted. Waiting for the official pitch deck means relying on stale metrics.

Auditing Misses in January and June

Because feedback takes half a decade to crystallize, investment teams easily drift into confirmation bias. Winners get celebrated while passed deals fade from memory. To stop that drift, Adams Street institutionalized a formal post-mortem on deals they declined.

“At Adams Street we have three to four times a year offsites and mini offsites, two main ones, one in Jan, one in June with the entire senior team to talk about a number of things including what we call our anti portfolio,” (play from 42:09) Jeevarathnam said. “What we missed three years ago, five years ago, and what was our underwriting then and how has it changed.” (play from 42:23)

Reviewing decisions three to five years after the fact removes real-time defensiveness. The market has delivered an objective verdict on the underlying company. By comparing original underwriting memos against actual operational outcomes, the team identifies blind spots in its selection criteria before those blind spots ruin future returns.

Parsing Proprietary Records With Machine Learning

Venture investing runs on private signals rather than public filings. Success depends on proprietary deal flow and non-public performance data gathered over decades. Adams Street tracks these records across thousands of fund partnerships and direct investments.

“Venture is an information asymmetry business,” (play from 0:00) Jeevarathnam observed. “We have confidential information within our core walls that we only share with ourselves, the investment team, that we can leverage to get to the best companies one or two steps earlier.” (play from 0:01)

The challenge is finding patterns buried across decades of PDF investor letters, cap tables, and board updates. Adams Street deployed machine learning models to search across these unstructured repositories. “This data is sitting in very disparate parts, unstructured data,” (play from 49:52) Jeevarathnam explained. “Sometimes we're using AI to help us connect those dots.” (play from 49:59) Connecting those dots lets the firm trace which founders and emerging managers repeatedly capture the 7% of deals that return the entire market.

Why It Matters

Power-law dispersion in venture capital is widening as fund counts swell past 4,000 vehicles. When 7% of startups drive aggregate performance, traditional point-in-time LP diligence fails to separate repeatable access from vintage luck. Sophisticated institutional capital is shifting toward systematic, long-cycle surveillance: tracking manager portfolio execution in real time and continually pressure-testing historical misses.