A fundraising benchmark only counts who already raised

The record

Carta's State of Private Markets: Q1 2026, published 29 May 2026, states its figures as activity among companies on Carta, reporting, for example, that more than 60% of venture capital raised by companies on its platform in the quarter went to AI companies. Crunchbase's methodology page, retrieved 16 September 2026, states that its data comes from aggregated sources plus voluntary user contributions, a combination the page itself says can leave successful companies overrepresented. PitchBook's report methodologies page describes a separate estimation step added specifically because private deals often surface only well after a transaction takes place. Each provider is describing where its own benchmark numbers come from before a founder ever sees a median.

What the sources establish

All three pages describe a dataset built from companies that completed a transaction the provider could see: a round closed on Carta's platform, a round reported to Crunchbase, or a round PitchBook eventually captured and, where necessary, estimated around. None describes a benchmark built from companies that tried to raise and failed, took longer than expected, or raised on terms materially different from what was announced. A median round size or valuation is a median of the visible, completed, disclosed cases, not of all attempts. Crunchbase's own language about overrepresentation of successful companies names this directly for its dataset; the same structural pattern applies to any provider that only sees a company once it has already closed something.

Scope and revision

The scope of a benchmark is the population that generated it, and that population is not all founders raising at a given stage. It is closer to founders who raised, at the stage the provider recorded, within the window the provider had visibility into. PitchBook's estimation step exists precisely because that window is incomplete at the time a quarter's benchmark is first published, and the figure can move once late-reported deals arrive. A benchmark drawn early in a quarter, before laggard reports surface, will understate activity in ways none of the three pages can fully correct for at the time of first publication.

The decision in front of you

This is an editorial checklist, not a substitute for legal or financial advice. Before treating a published median as a target, ask what population produced it and how completely that population was captured.

• Does the benchmark's source state whether it includes companies that did not close a round?
• Is the reporting window for this figure still open to late-arriving deals?
• Does the sample come from one platform's customers, one aggregator's disclosures, or a broader capture method?

A benchmark tells a founder what happened among the deals a provider could see. It is a floor on what is knowable, not a ceiling on what is normal.