Database methodology

How the database is built.

Every row in the Institute’s database can be traced to the document it came from. This page explains where the records come from, how filings become firms, what each evidence label means, and the limits to keep in mind when you use it.

765,661SEC Form D filings read, from 2009 to today
5,967Venture firm families
98,842Companies, with how each is known
242Limited partners, and growing


Colibrí Institute VC Database, counts as of September 21, 2026. The database updates every hour, so the live figures will differ.

What the database is

The database is an open record of the United States venture capital market, built from documents anyone can check. It is organized around five questions people ask when they raise or deploy capital: which firms back companies like mine, how a firm compares with its peers, what a firm has invested in, who sits on a company’s board, and which limited partners back funds like mine.

Each of those questions is a view in the database. Every row can be opened to show the evidence behind it and a link to the filing, report or page it was read from. You can search, filter and sort any view, and export exactly what is on screen.

The database is separate from the research behind our white papers. Those papers analyze a fund-level performance sample licensed from PitchBook as part of doctoral research. The database is collected independently from regulatory filings and public statements, and it stores no licensed vendor data. Figures from one should not be read as figures from the other.

Sources

SEC filings, limited partners’ own published reports, and what firms and allocators publish on their own websites. Each is labelled wherever it appears.

Updates

New filings are collected daily and the public views refresh every hour.

Reuse

The database is free to use. Please cite it with the date you accessed it.

The database

Search it, filter it and export what you need.

Open the database

Sources

Where the records come from

Most of the database comes from two filings that private funds and companies make with the Securities and Exchange Commission. The rest comes from limited partners’ own reports and from what firms and allocators say about themselves on their own sites.

SEC Form D
The notice a fund or company files when it sells securities privately. A fund’s Form D gives its size, how much it has sold, how many investors it has admitted, its minimum investment and the people who sign it. A company’s Form D describes a financing round and names its directors. We read every Form D filed since January 2009, with every amendment.
SEC Form ADV
The adviser registration filed by investment advisers, including the exempt reporting advisers most emerging venture managers file as. It gives assets under management, fund counts and, uniquely among filings, the firm’s own website. We read the full monthly roster.
Limited partners’ reports
Public pensions are required by law to publish their private markets holdings. We read these reports directly for the commitments they record: which fund, how much and which vintage.
Firms’ own websites
What a venture firm publishes about itself: the companies it names in its portfolio, its team and partners. Kept word for word with the page and the date we read it, and always labelled as the firm’s own statement.
Limited partners’ own websites
What an allocator says it invests in, how it defines an emerging manager, how it asks to be approached, and the managers it names as ones it backs, quoted with the page it came from.
Press coverage
News headlines about limited partners’ fund commitments. Press is listed separately as a headline and a link. It helps us find allocators to read, and it never adds to a count.
Public rosters
Lists published by organizations that name women and underrepresented general partners. A roster entry is that organization’s statement. It reaches a firm only when the person it names appears on the firm’s own filings, and only after a person on our team confirms the match.

From filings to firms

Filings arrive one legal entity at a time, and a single manager may file dozens of vehicles under different names. Before the database can say anything about a firm, it has to decide which filings belong together and which ones describe a fund at all.

Funds, not deal vehicles
Every issuer is classified once as a fund, a special purpose vehicle, an operating company or another kind of entity. Special purpose vehicles, which raise money for a single deal, are left out of every fund figure: vehicle counts, filing counts and capital raised count pooled funds only.
Families
A manager’s vehicles are grouped into one family by a consistent rule applied to their names, and a parent’s vehicles are grouped under it only when the same small group of people signs them. Records are never merged because they look alike; a possible match waits for a person to confirm it.
Venture firms
The Firms view lists families whose filings identify them as venture managers, or whose own site does where the filings do not say. Wealth managers running a venture vehicle for clients are left out.

Emerging

The Firms view applies the Institute’s published definition. A firm is emerging if it meets any one of four criteria: less than $100 million in assets under management; a current fund that is Fund I, II or III; a management firm less than five years old; or a founding team that includes women or otherwise underrepresented general partners.

How each criterion is read

Assets come from Form ADV where the manager files one. Fund sequence is not counted for a side vehicle raised inside an established firm, while a true spin-out keeps it. Firm age is dated from the first sale of the firm’s first fund. The fourth criterion comes only from a partner’s own public statement or a confirmed roster entry. A blank means the evidence has not been found, not that it is absent.

Activity

Raising: a fund is on file that has not yet reported a first sale. Deploying: a fund filed within two years. Quiet: the last fund filing is two to five years old. Dormant: no fund filing in five years.

Definitions

The emerging manager definition is explained in full with the research.

Read data and methods

Evidence

Two kinds of evidence, never blended

A company tied to a firm by a filing and a company a firm names on its own website are different kinds of fact. Both are useful, so the database keeps both, labels every row with the kind it is, and never adds them together into one count.

Filed

Read from a document filed with a regulator or published by a limited partner as required by law. A filed row carries the figures that document gives, such as dates, amounts and investor counts.

Examples: a company’s own Form D; a pension’s published commitments.

Stated

What a firm or an allocator says about itself on its own website, quoted with the page it came from. A stated row has no filing behind it, so the filing columns stay empty.

Examples: a firm’s portfolio page; an allocator’s emerging manager programme page.

Evidence labels · What each one means

LabelWhere you see itWhat it means
Filed with the SECCompaniesEvery figure comes from the company’s own Form D filings.
Named by a firmCompaniesA venture firm names the company on its own site. No filing ties them.
City and state matchDeals, Board seatsThe director on the company’s filing matches the manager’s signatory on city and state, and is a director rather than an executive.
State matchesDeals, Board seatsThe same match, on state only. Weaker than a city and state match.
Firm’s site confirmsDeals, Board seatsThe filing match holds, and the firm’s own page also names the company.
Firm site, company found in filingsDealsThe firm names a company that has filed a Form D of its own.
Firm site, links the companyDealsThe firm’s page links to the company’s own website.
Firm site, company pageDealsThe firm’s site has a page about the company.
Firm site, named on its portfolio pageDealsThe firm names the company on a portfolio listing.
The limited partner’s own published reportLimited partnersCommitments with amounts and dates, read from the allocator’s report.
The limited partner’s own siteLimited partnersWhat the allocator says about itself and the managers it names.
The manager’s own siteLimited partnersA manager names this allocator as a backer.
Named in press coverageLimited partnersA news headline names this allocator. Its own site has not been read yet.

Portfolios and board seats

No filing lists a venture fund’s portfolio. The database reconstructs one in two ways and shows both.

From filings
When a person who signs a manager’s fund filing is also named as a director on a company’s Form D, we record a board seat. A deal is one firm, one company and one board member. Its entry date is the company raise closest to the seat, and follow-on figures count only raises filed after it.
From the firm’s own site
Firms publish far more of their portfolio than filings reveal. A name on a firm’s site counts only with evidence beyond the words themselves: it links to the company’s own site, has a page of its own, appears in the filings, is named by other firms as a company, or sits on a genuine portfolio listing. Menu items, section headings, legal pages and link text are set aside.
How companies are named
A company is shown by its name, taken from the firm’s page when it reads as one and otherwise from the company’s own site. When several vehicles of one firm carry the same page, the company appears once, under the firm.

Counts are floors

Filings tie only about two companies to each firm on average, because most portfolio companies never file a Form D naming a partner as director. The firm-site count fills much of that gap, but the rule that removes menu words also misses some real names. Both counts are floors.

No dates on stated rows

A website rarely says when an investment was made or how large it was, so firm-site rows carry no dates or amounts.

Limited partners

The Limited partners view is built for a manager raising a fund. One row is one allocator. Opening a row shows who else it has backed, with a link to each manager’s site, what it says it invests in, how it defines an emerging manager and how it asks to be approached, each in its own words.

Amounts and dates
These come only from a limited partner’s own published report. Public pensions publish commitments by law, which is why they carry the most detail.
Recent activity
An allocator that has not committed to a fund in three years is flagged, because it is unlikely to be a prospect for a fund raising now.
Allocators only
Placement agents, accelerators and industry networks are listed because managers meet capital through them. They do not commit capital themselves, so they are hidden until you turn off the Allocators only filter.
New allocators
The list grows as allocators are found in reports, on managers’ sites and in press coverage. An allocator found only in press is labelled that way until its own site is read, and a name that may duplicate one already listed waits for a person to review it.

What the database leaves out, on purpose

Licensed vendor data
No data from PitchBook, Crunchbase, Dealroom, Tracxn, CB Insights or LinkedIn is stored. The block is enforced in the database itself.
Returns reported by managers
No fund’s own return, multiple or valuation, as its manager reports it, is ever shown in the public database, and none enters the Colibrí 100. Managers’ returns count only inside combined benchmark figures of at least five firms, in their own table labelled as reported by managers, and in the Institute’s aggregate research. A manager sees its own funds’ figures on its private benchmark card.
Inferred demographics
Nothing about a person’s gender, race or ethnicity is ever inferred from a name or a photograph. It is recorded only from that person’s own public statement or a confirmed public roster.
Automatic merges
Two records are not combined into one firm or one allocator without a person confirming it. The one exception is fund families that share a website of their own, which are treated as one firm. We would rather show a firm twice than merge two firms that are different.
Rankings
The database describes firms and allocators. It does not rank or rate them.
Guesses
A figure we cannot trace to a source is left empty rather than estimated or filled with an average.

Limitations

Keep these in mind when you use a figure, and quote them with it.

Form D is a notice
Amounts are as filed and may change by amendment. Many issuers never file a closing amendment, and some offerings never require a Form D at all.
Coverage is uneven
About half of venture firms have a website on file, and some managers are missing from the Firms view because their filings do not identify them as venture managers.
Deal vehicles and small funds
The line between a special purpose vehicle and a small pooled fund still relies partly on offering size. Some small funds may be classed as deal vehicles, and the reverse.
The fourth criterion
Evidence on women and underrepresented general partners covers a minority of firms so far. Most firms’ status on that criterion is unknown rather than negative.
Limited partner coverage
Only a small number of public institutions publish reports detailed enough to read commitments from them. Most allocators never publish who they back.
United States funds
Firms and companies are drawn from U.S. filings. Allocators outside the United States appear only where they back U.S. managers or appear in coverage of them.

Names

Some firm rows carry the name of a fund vehicle rather than the firm, and some companies appear under their web address when their site gives no usable name.

Websites on file

A small number of firms have a website on file that we believe belongs to another organization. These are held for review, and their stated figures should be read with care.

Benchmark

The Fund Benchmark

The Colibrí Fund Benchmark answers one question for a manager or an allocator: is this fund normal for its cohort, and if not, in which direction? It is free, published under the same CC BY 4.0 license as the rest of the database, and every figure names its source and how many funds stand behind it.

Filed

Fundraising and structure from each fund’s own Form D and its adviser’s Form ADV: stated target, share of target on file, amount sold, investors on the filing, minimum investment, months between first and latest filing, whether the adviser names an auditor and an administrator, and how often a Fund I or Fund II is followed by the next numbered fund within four years.

Label on the page: Filed.

LP-published

Net IRR, net multiple and DPI by vintage band, taken from public pensions’ and endowments’ own published reports about funds they hold, net of fees. Each institution is named with its report date and a link to the document.

Label on the page: LP-published.

Stated by managers
Terms no filing carries, such as management fee, carried interest, GP commitment, recycling, fund term, planned positions, check range, reserves, target ownership and time to final close, as managers give them through Update your record. They are combined anonymously with the cohort only with the manager’s consent and only once five firms in that cohort have added theirs, and they sit in their own column, never beside a filed or LP-published figure.
Reported by managers: net returns
Net IRR, TVPI and DPI that managers give us for each of their funds; a firm with several funds adds each one. Limited partners that publish returns mostly hold larger, established funds, so this is the benchmark for the emerging and smaller funds they rarely report on. These figures are combined anonymously by vintage band, sequence and target, publish a median at five firms, quartiles at twelve funds and the 10th and 90th percentiles at twenty, and no single firm may be more than a fifth of any figure. They sit in their own table, labelled as reported by managers, and are never mixed with LP-published returns. No fund’s own figure is ever shown publicly, and none enters the Colibrí 100 score.
Cohorts
A fund’s cohort is its sequence (Fund I, Fund II, Fund III, Fund IV, Fund V and later), its stated target band (under $10M, $10M to $25M, $25M to $50M, $50M to $100M, $100M to $250M, $250M to $500M, $500M and over), the three-year band in which it first filed, and optionally its region, including all US regions together. Three-year bands match the way limited partners group vintages and keep enough firms in each cohort for a figure to publish. Only pooled venture funds count; special purpose vehicles are left out, a fund filed through several parallel entities counts once, and each fund counts at its latest filing. LP-published returns are shown by vintage band only, because limited partners report funds by name and vintage.
Filters: sector, stage focus, manager and adviser
A biotech fund is not an agriculture fund, so filed fundraising and the share of firms that raise a next fund can also be read by up to two filters at a time, alongside sequence, target and first-filed years. Sector comes from one of two bases that are never merged. As a firm states it: the sectors a firm names on its own site, where a firm that names several counts in each and a firm that names none is left out rather than guessed. In its filed portfolio: at least half of the companies tied to the firm by SEC filings name that industry on their own Form D, which covers fewer firms but rests on nothing the firm says about itself. Stage focus is what a firm states on its own site. Manager follows the Institute’s published emerging manager definition. Adviser is the firm’s Form ADV regime on file: exempt reporting, registered, or none found. The page shows how many firms carry each filter next to it. A filtered cohort is stored only once it has five firms and follows the same publishing rules as every other figure; returns and terms are shown for the whole cohort until enough limited partners and managers report by these filters.
When a figure publishes
A filed figure publishes a median at five firms, quartiles at twelve and the 10th and 90th percentiles at twenty, and no single firm may be more than a fifth of it. An LP-published return figure needs five funds from at least two institutions, because a figure fed by one institution is that institution’s portfolio; its quartiles need twelve funds and its 10th and 90th percentiles twenty. A figure that falls short shows its count and the reason, and the page offers the nearest cohort that publishes, one dimension wider, naming what it widened.
Quarterly releases
The benchmark is released quarterly. The current quarter updates daily and becomes final on the last day of the quarter. A final quarter is a citable record and is never rewritten; a correction appears in the next quarter. The first release is Q3 2026, final on 30 September 2026.
Caveats beside the number
Amount sold and share of target are floors, because many funds never file a closing amendment. Investors on the filing are subscribers counted by the issuer, not limited partners by any institutional definition. Net IRR for young vintages annualises over a short holding period and reads high before distributions begin, so the page says so on the row. Each caveat sits next to the figure it qualifies.

The private card

Every firm has a private benchmark card built from its filings alone, placing its latest fund’s figures against its cohort as percentiles, and each fund’s net returns against both the LP-published and the manager-reported benchmark, with a PDF for a data room. A sample card shows a fictional fund on the live figures. A sign-in link goes at once to an email at the firm’s own web address; any other address waits for a person to approve it. The card is private. A firm may choose to show its position on its public record, labelled as computed by the Institute and shown by the firm’s choice.

Notices

No firm is emailed because it appears in a filing. A contact who has confirmed an email with us receives at most one notice per cohort, when that cohort is one manager short of publishing, and every notice carries a one-click stop. Anyone can ask the benchmark page to tell them when a cohort publishes.

What is never public

A named fund’s stated terms unless its manager chose to show them, a firm’s percentile it did not choose to show, any fund’s own reported returns (they count toward the combined figures and the Institute’s research, but are never shown on their own), a rank, and any figure that falls short shown as a number rather than as the gap it is.

Widening the return figures

Managers can name limited partners that publish fund-level performance. We read those reports ourselves, and anything taken from them enters as LP-published with a link to the document, never attributed to the manager.

Open the benchmark

Corrections and citation

Corrections

If a row is wrong, tell us and send the source. We correct it, date the correction, and keep the record of what changed, following the Institute’s corrections policy. A firm’s place in the database is not open to negotiation, only to evidence.

Citing the database

The database changes every hour, so name the view, any filters you used and the date you accessed it. Suggested citation: Colibrí Institute. Colibrí Institute VC Database. Accessed [date]. www.colibri.institute/database.

Version

Methodology version 1.3, September 26, 2026. Adds sector, stage focus, manager and adviser filters to the Fund Benchmark. Version 1.2, September 25, 2026, added the Fund Benchmark (section 10), including net returns as managers report them, combined anonymously and never shown fund by fund; finer target bands and fund sequences; a US-only region; and the 10th and 90th percentiles. Material changes to a rule will be published here with a note of what changed.

Questions about method

hello@colibri.institute