The national figure describes the economy, not your contact list. A venture team's network is a selected sample: founders know founders, and partners meet companies that pitched them. That selection is the reason the question needs its own study instead of a borrowed statistic.
Known, conditional and suspected
| Type | Statement | Basis |
|---|---|---|
| Known | About 0.11% of new U.S. companies created in 1981 to 2005 received venture capital | Puri and Zarutskie's Census-based study, as reported in an NBER survey of venture capital research |
| Known | Venture capitalists typically invest largely in young firms, with potential for large scale an important criterion | Puri and Zarutskie, NBER working paper 14250 |
| Program fact | Introduction requires 50+ full-time employees at the company's peak (contractors never count), an operating history it can document, rights to license and someone authorized to act | SourceX terms and who-qualifies page, checked 2026-10-26 |
| Conditional | Whether any counted company is accepted, licenses records or produces a reward | Decided only in SourceX review and a signed agreement |
| Suspected | In an early-stage fund's reachable network, most companies that pass an operating-history screen sit outside the portfolio | Our hypothesis; no fund has been measured |
The one question the study answers
Among companies a fund's team can reach through a named relationship owner, what share of those that pass an operating-history screen sit outside the current portfolio?
Every word in that sentence is a measurement decision, so each one is pinned down before any company is listed.
| Term | Definition used in this study |
|---|---|
| Reachable company | A company whose CEO, or a direct report of the CEO, someone on the team or an LP contact can name and ask for a conversation |
| Relationship owner | The person who would make that ask; recorded by role, never by name in outputs |
| Inside | A current portfolio company of the fund on the frame date |
| Outside | Everything else: exited or written-off holdings, LP-owned operating companies, portfolio customers and suppliers, companies where a partner holds an outside board seat, employers of platform-program members |
| Operating-history screen | 50+ full-time employees at peak, at least five years in operation, and records kept in a system the company names; the five-year line is this study's choice, not a program rule |
| Unknown | Any screen fact the relationship owner cannot state; a category in its own right, never silently treated as a pass or a fail |
A parent company and its subsidiaries are counted once, which matches how the program applies the reward cap and stops a holding group from inflating the outside column.
The network-composition study design
The output is a four-cell table plus an unknown row. Cell a holds inside companies that pass the screen, b outside companies that pass, c inside companies that fail and d outside companies that fail. The primary estimate is b divided by a plus b: the outside share of screen-passers. The graphic shows the layout with every cell left empty.
| Protocol element | Rule |
|---|---|
| Sampling frame | Five lists merged on one frame date: current portfolio, exited and written-off holdings, the CRM at company level, LP-disclosed operating companies, and outside board seats and platform members |
| Cohort dates | Frame frozen on day 0; screen facts collected by day 30; the same frame re-read on day 180 to see which unknowns resolved |
| Unit | The company, grouped by parent; one row each, whoever knows it |
| Consent, internal | Each relationship owner agrees before their company is recorded, and may withdraw a row until results are aggregated |
| Consent, external | No company is contacted for the study. Facts come from what the relationship owner already knows or from public filings. A company hears about licensing only if its CEO wants that conversation, which is separate from the study |
| Variables | Inside or outside, outside sub-type, founding year, peak full-time band (under 50, 50 to 199, 200 to 999, 1,000 or more), named record system yes or no, years of records band, rights picture (company-owned, client-owned, unknown), relationship owner role, status |
| Status codes | Not asked, owner declined to ask, CEO declined, CEO agreed, introduced, not a fit, in review, accepted |
| Missing data | Report the outside share three ways: unknowns as fails (lower bound), unknowns excluded, unknowns as passes (upper bound) |
| Anonymization | Bands, never exact headcounts; no names; any published cell under five is suppressed and merged, following the k-anonymity idea NIST describes, in which every combination of identifying attributes matches at least k records |
| Pre-registration | Date-stamp this protocol before the frame is built; log any later change as a deviation with its date and reason |
| Analysis | Descriptive only: the outside share with a binomial interval, the share by outside sub-type, and status counts by column |
Methods note. The design measures where reachable, screen-passing companies sit and how far each relationship got. It draws on the program's published fit criteria (SourceX terms and who-qualifies page) and on the Census-based venture research cited above, all checked 2026-10-26. It does not measure demand for any company's records, record quality, rights in detail, payout or any reward; those are assessed only in SourceX's review. For listing the network in the first place, the network opportunity worksheet and the network screening guide give the columns.
Why the companies that say no stay in the table
A composition study that keeps only successes answers a different, flattering question. Every company named on day 0 stays in its cell to the end, whatever happens to it.
| Status at day 180 | Stays in the table as | What it does not tell you |
|---|---|---|
| Owner declined to ask | A reachable company with no ask made | That the company would have said no |
| CEO declined | A company that heard the request and passed | That the records were unsuitable |
| Not a fit after review | A company that failed a test the screen missed | Which test, unless the reason is recorded |
| In review | An open case, reported separately | Anything about the outcome |
| No relationship owner after all | A frame error: moved to a separate count, not deleted | Whether the company passes |
Two cautions on reading the result. First, a high outside share is a description, not a cause: outside companies are older on average for reasons that have nothing to do with the fund, and being outside does not make a company fit. Second, if outside companies are introduced more often, that may reflect who owns those relationships rather than the companies themselves. The conversion denominator guide and the partner quality study cover the same traps for later stages.
How four rows are coded (Illustrative)
Illustrative. Generic companies, invented facts, no results implied.
- A 31-year-old regional distributor owned by an LP's family office: 260 full-time at peak, ERP records since 2009. Outside, LP-owned, passes the screen. The principal asked; the CEO declined until an acquisition closes. It stays in cell b as "CEO declined".
- A Series B portfolio company founded in 2019: 140 full-time at peak, but its richest records sit inside customer tenants. Inside, passes the screen, rights picture "client-owned". Passing the history screen is not fit, and the who-qualifies page treats client-owned material as usually not a fit.
- A former portfolio company absorbed in 2023 into a 900-person group whose CEO nobody knows. Outside, exited, passes the screen, no relationship owner. Counted, not dropped: the row shows a relationship the network once had and lost.
- A 2024 seed company with 30 staff. Inside, fails the screen. It sits in cell c with the condition that would move it: full-time headcount passing 50.
Row 4 is the counterexample worth keeping in mind. A fund whose portfolio is all rows like it will find a high outside share almost by construction; that tells you where to look, not that outside companies are better.
What would make the thesis wrong
- Fund stage. Puri and Zarutskie find venture capitalists invest largely in young firms, which is what the thesis rests on. A growth-stage or buyout-heavy fund holds older, larger companies, and its inside column could dominate.
- Frame undercoverage. Teams list the companies they think of. Outside companies with quiet relationships, such as an LP's legacy business, are the easiest to forget, so the frame can understate b. Merging five lists is the control, not a cure.
- Rights. Operating history is not licensable history. The who-qualifies page says material belonging to clients is usually not a fit, and that is common in portfolio software companies.
- Demand. SourceX checks relevance to its current data needs in review, and the page adds that a company out of scope for one program may come into scope for a later one. A strong screen result can still meet no current demand.
- Missing not at random. Confidential LP relationships are the likeliest to be withheld from the frame, and they are disproportionately outside companies. The three bounded shares make the effect visible but cannot remove it.
Run the four-cell frame now and re-read it at day 180
Build the frame in one working session of about 40 companies, collect screen facts within 30 days and report the four cells, the unknown row and the three bounded shares. Re-read the same frame at day 180. The recommendation changes on one number: if the inside share of screen-passers is half or more, start with the portfolio and treat outer relationships as second priority. Whoever submits an introduction must first sign the SourceX Channel Partner Agreement, formerly titled the Channel Partner Non-Circumvention and Non-Disclosure Agreement, and when several people introduce the same company, only the first valid introduction inside the 90-day attribution window earns credit, which attaches once the company submits a verified application. The overview for venture platform teams covers the portfolio side.
What a counted company still has to show
The study screens at metadata level only: record type, rough counts, span of years, the system and who administers it. Nobody asks for files, samples or logins. Before an introduction, a company still needs the full published fit. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. The CEO authorizes a licensing discussion and signs for the company; a fund's shareholding does not make it the decision-maker. Three answers wait for SourceX's own review: whether the records match current demand, whether customer contracts limit reuse, and what any license would cover. The metadata-only opportunity map lists the questions.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
When a row in cell b has a willing relationship owner and a CEO open to talking, register as a partner, set that company beside the who-qualifies criteria and read the terms of the program before anyone makes the ask.
Running this study is network research, not diligence on SourceX; readers considering SourceX as an investment can request investor materials through the SourceX investor overview, where the company says it shares company and commercial metrics with investors directly.
This article is for information only and is not an offer to sell or a solicitation of an offer to buy securities. Any offering by SourceX is made only to verified accredited investors through its offering documents.