The decision for a platform research director is whether to commission that study for your network, and how to read anyone else's claim that a network "represents" data supply. Funding status never establishes eligibility, so the study ignores it.
Known, conditional and proposed
| Statement | Standing | Source |
|---|---|---|
| An introduced company is screened on reaching 50+ full-time employees at its peak (contractors do not count), operating history on record, rights to license and an authorized contact; how it was funded never enters the screen | Known program fact | SourceX terms and who-qualifies page, checked 2026-10-26 |
| An introduction made without the decision maker's knowledge, or for material that belongs to the company's clients, is usually not a fit | Known screening guidance | Who-qualifies page |
| Whether a sampled company is accepted, what it licenses and whether a reward follows | Conditional | SourceX review and a signed agreement |
| A fund's non-portfolio relationships hold more companies that pass the size, history and sponsor tests than its portfolio does | Proposed thesis | Unmeasured; tested by the protocol below |
| Term | Meaning here |
|---|---|
| Relationship stratum | One of five relationship types, sampled separately |
| Apparent opportunity | A company in the network that looks like a fit from the outside |
| Permissioned opportunity | An apparent opportunity whose relationship owner agreed to ask and whose CEO knows about and wants the conversation |
| Sampling frame | The complete, dated list the sample is drawn from |
| Elimination ledger | The record of every sampled company that dropped out, with the stage and the reason |
The business owners absent from an AI startup market map
They run companies no technology category describes: regional distributors, claims administrators, field-service groups, specialty contractors.
Venture portfolios are organized by technology category. Andreessen Horowitz's About page, for example, says the firm invests in seed, venture and growth-stage technology companies across AI, bio and healthcare, consumer, crypto, enterprise, fintech, games, infrastructure and American dynamism. A market map built from portfolios like that has no box for a 28-year-old distributor, even when its owner is an LP, sits on a portfolio company's customer council or employs a former portfolio founder as COO.
Platform networks are built to reach past the cap table. The same firm's enterprise page says its operating team gives entrepreneurs access to a robust network of experts. So the network a fund can reach is not the network its market map shows. Neither is evidence of data supply. That gap is the research question: which of the companies a fund can reach, but does not own, have the records, the rights and an authorized sponsor?
Map each relationship type as its own stratum
Each type reaches a different kind of company through a different person, so a pooled list hides which relationships produce fit.
| Stratum | Who sits in it | Who owns the relationship | Evidence quality | Bias to expect |
|---|---|---|---|---|
| Portfolio | Current holdings | The founder-CEO, through the deal partner | High: dated board materials | Young companies, often below 50 full-time employees at peak |
| Alumni | Companies now run or staffed by former portfolio founders and executives | The alumni themselves | Medium: what they report about their new employer | Alumni are often functional executives, not CEOs |
| Customer | Customers and suppliers of portfolio companies | The portfolio founder who holds the account | Low: secondhand | Founders protect accounts, so consent may be scarce |
| Advisor | Employers and clients of experts in the platform network | The expert | Low to medium | Experts may describe client records they do not own |
| Owner-led | Companies whose owner-CEO the fund knows directly: LPs, co-investors, board peers | The partner or LP principal | High: owners know their own companies | Small stratum; confidential relationships |
Building the frame. Export companies, not people, from the fund's CRM, platform databases and expert network. Deduplicate. Assign each company to one stratum by its strongest relationship, using a tie-break written before sampling (for example owner-led, then portfolio, alumni, advisor, customer). Date the frame and publish the count per stratum.
Drawing the sample. Draw the same number of companies from each stratum at random, for example 12, so strata can be compared. Weight back to the frame counts only when estimating for the whole network.
This differs from the two screening tools already on the site. The five-ring map for platform leads ranks which conversations to have next, and the network screening guide walks one person through everyone they know. The protocol here measures the network before anyone asks anything.
Which sampled companies have operating history and an authorized sponsor?
Code each sampled company on ten variables, each dated and tied to a source type, from what its relationship owner already knows. Nobody contacts the company to fill in the codebook.
| Variable | Values | Allowed source |
|---|---|---|
| 1. Stratum | Portfolio, alumni, customer, advisor, owner-led | The frame |
| 2. Relationship owner | Role of the person who could ask the CEO | Fund records |
| 3. Peak full-time headcount | Under 50, 50 to 199, 200 or more, unknown | Board materials, owner statement, public filing |
| 4. Years of own operating records | Under 3, 3 to 10, over 10, unknown | Relationship owner |
| 5. Record system class | ERP, CRM, helpdesk, maintenance, claims, other, unknown | Relationship owner |
| 6. Ownership flag | Company-owned, client-owned, mixed, unknown | Relationship owner; never a contract review |
| 7. Location and record language | Qualifying, needs an English check, outside, unknown | Relationship owner or public record |
| 8. CEO path | Direct, through one person, none, unknown | Relationship owner |
| 9. Permission status | Agreed to ask, declined, not asked | Relationship owner |
| 10. Evidence date | Date of the newest fact in the row | All of the above |
Funding stage, valuation and growth rate are deliberately absent. None of them bears on eligibility, and recording them invites the very confusion the study exists to remove.
Where do permission and fit eliminate apparent opportunities?
At five stages, in a fixed order, with every elimination kept in the ledger with its reason.
| Stage | Elimination reason | Kept in the denominator? | What it tells you |
|---|---|---|---|
| A | Relationship owner declines to have the row counted | Yes, as "declined to be counted" | How guarded that relationship type is |
| B | A published fit fact fails: headcount at peak, years of records or location | Yes | Whether the stratum holds established companies at all |
| C | Records belong mostly to the company's clients | Yes | Whether visible data is actually licensable by the company |
| D | No path to the CEO | Yes | Whether the relationship reaches decision-makers |
| E | Relationship owner declines to ask, or the CEO does not want the conversation | Yes | The real cost of permission |
The headline figure per stratum is permissioned opportunities divided by companies sampled. Rows still unknown at the cutoff are reported as a range: the lower bound treats them as eliminated, the upper bound as passing. This is not a conversion rate. Nothing in the study observes SourceX's review, and a later study of completed handoffs belongs to the conversion benchmark design. How often each path reaches the CEO is measured in which introductions actually reach an authorized data owner.
Illustrative row, alumni stratum. A former portfolio CFO is now COO of a 300-person industrial distributor with 17 years of order and returns history in one ERP. Headcount, records, system and location are known, and the COO describes the records as company-owned. The CEO path runs through the COO. The alumna agrees to ask; the COO raises it with the CEO, who wants a conversation. The row ends as a permissioned opportunity, dated. Everything after that is SourceX's review, outside the study.
Illustrative no-fit, customer stratum. A portfolio company's largest customer is a 500-person payroll outsourcer. Size and history pass easily. Nearly every record it holds describes its clients' employees, so the row ends at stage C and stays in the denominator. On a pooled list it would have looked like one of the network's most "data-rich" companies.
The network opportunity map, with dated evidence
One table per study wave: strata as rows, stages as columns, counts in the cells and the cutoff date in the title. Until the study runs, every cell reads "not yet collected".
| Stratum | Sampled | Fit facts known | Passing fit facts | CEO path | Permissioned |
|---|---|---|---|---|---|
| Portfolio | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| Alumni | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| Customer | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| Advisor | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| Owner-led | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
Four rules keep the map honest. Every fact carries its date, and facts older than 12 months are flagged stale. Unknowns are reported, never filled with guesses. A later wave re-reads the same rows instead of topping up with new companies. Company names never leave the fund: the published map shows counts only, after an anonymization check that no cell identifies a company.
Method note. The strata, codebook and elimination stages are built from SourceX's published fit criteria, screening guidance and attribution rule (terms and who-qualifies page, checked 2026-10-26) and from public descriptions of how venture firms organize portfolios and platform networks. The study measures relationship paths and known fit facts on a stated date. It does not measure demand for any company's records, record quality, the value of a license or any payout, which only SourceX's review and a signed agreement establish.
What would show this network thesis is wrong
Write these decision rules before collecting a single row.
- The portfolio is enough. If the portfolio stratum yields at least as many permissioned opportunities per company sampled as the four outer strata combined, the hidden network adds little for this fund.
- The network is not really known. If more than half of sampled rows stay unknown on headcount or records after asking the relationship owner, the fund cannot claim anything about its network yet.
- Ownership dominates. If most rows that pass size and history fail at stage C, the visible data belongs to other people.
- Permission dominates. If relationship owners decline to ask for most rows, the network exists on paper only.
Risks and limits of the thesis
- Rights. The most visible records may belong to clients, and SourceX's screening guidance treats client-owned material as usually not a fit.
- Demand. SourceX checks relevance to its current data needs in review, and its programs differ by company size, operating history, geography, language and data category. A permissioned row can still be declined.
- Measurement. Relationship owners report facts secondhand, and errors grow in the customer and advisor strata. Have a second coder recode a subset and report agreement.
- Selection. Relationship owners who agree to be counted may be those with the strongest relationships, which flatters the results.
- Substitutes. If synthetic data or public sources cover more of what AI developers need, demand for company records could narrow. This study does not measure that.
- Borrowed benchmarks. Partner-channel figures from other markets are mostly self-reported vendor surveys. PartnerStack and Wynter's 2026 survey of 100 senior B2B SaaS leaders, for instance, reports that mid-market and enterprise companies count 35% of pipeline as partner-influenced or sourced. Useful context about software sales; no evidence about any fund's data network.
A measurable next step. Draw 12 companies per stratum, 60 rows in all, and code them in one session with the relationship owners. Publish the map with its cutoff date, then re-read the same 60 rows six months later.
Before any sampled company is introduced
The study ends where an introduction would begin, and introductions follow the program's own rules.
- Fit. A full-time workforce that reached at least 50 at its peak, leaving contractors out; years of operation the company can evidence; the right to license what it holds. 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.
- Authority. The CEO authorizes a licensing discussion and signs for the company. Sponsors and relationship owners carry the question; they do not decide it.
- Metadata only. Nobody in the fund asks for or handles a company's records, samples or credentials. Discovery stays at what, how much, how long, which system and who owns it.
- Credit. A sampled company can be credited to one partner only, and that is whoever made the first valid introduction, provided the company's verified application follows within the 90-day attribution window. Agree in writing whether the relationship owner or the platform team submits, and the submitter must first sign the SourceX Channel Partner Agreement.
What stays unknown until SourceX reviews a company: whether its records match current data needs, whether contracts limit reuse and what a license would cover. For the portfolio side of the program, see the overview for venture platform teams; for owner relationships held by LPs, see the family office page.
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 on your map becomes a permissioned opportunity, sign up as a partner, compare that company with the qualification criteria and go through the program terms before anyone makes the introduction. More research guides are in the guides hub.
Sampling a network to see whether it exists is a referral question, while evaluating SourceX as an investment is a different one; its investor materials can be requested through the SourceX investor overview.
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.