A list of the companies you know can only show you whom you know. It cannot show what is absent, which is the whole question. A denominator per cell turns "we don't know anyone in claims work" from an anecdote into a coverage figure you can compare across cells, track over time and act on.
Write the supply thesis as cells, not as a sector story
A supply thesis usually starts as a sentence. Ours, for this page: agents evaluated on back-office work need records of multi-step operational work that ends in a verifiable state. That sentence names no company. Split it into cells, each one an industry, the record the thesis needs from it and a size band:
- shipment exception logs at freight intermediaries;
- claim adjudication files at third-party administrators;
- work orders with closure codes at maintenance contractors;
- denial and appeal records at medical billing services;
- order corrections and returns at wholesale distributors.
Each cell gets one NAICS industry code before any counting, because the public firm counts are organized that way.
| Type | Statement | Basis |
|---|---|---|
| Public data fact | SUSB is an annual Census Bureau series covering establishments with paid employees by geography, industry and enterprise size: firms, establishments, employment in the week of March 12 and annual payroll; the latest data year listed is 2022 | Census Bureau SUSB pages, checked 2026-10-26 |
| Program fact | Introduced companies need 50+ full-time employees at peak, years of their own records, rights to license and an authorizing CEO | The who-qualifies page, read with program terms version 2026-10-22.2, since replaced by version 2026-11-04 |
| Conditional | Whether any cell's records match a current SourceX program, and whether a given company is accepted | Decided only in SourceX review |
| Our thesis | A network built around venture deals reaches operating industries unevenly, and its empty cells mark where supply is missing from its view | A working hypothesis, not measured; the test below checks it |
The network coverage gap matrix
Each row is one cell. The firm count is the denominator; the three network columns narrow from anyone you know to companies that pass the published fit facts on metadata.
| Cell (Illustrative) | Firms in the matched industry with 50+ employees (Illustrative) | Companies anyone in the network knows | With a CEO path | Passing fit facts on metadata | CEO paths per 1,000 firms | Reading |
|---|---|---|---|---|---|---|
| A. Shipment exception logs | 1,200 | 15 | 9 | 4 | 7.5 | Covered |
| B. Claim adjudication files | 300 | 0 | 0 | 0 | 0.0 | Thin market |
| C. Maintenance work orders | 2,500 | 1 | 0 | 0 | 0.0 | Empty network |
| D. Billing denial appeals | 800 | 8 | 6 | 0 | 7.5 | Ownership wall |
| E. Order corrections and returns | 1,600 | 14 | 2 | 1 | 1.3 | Wrong level |
Illustrative: the cells, firm counts and network counts are invented to show the arithmetic. They are not Census figures; replace every number with your own lookup and your own network records.
A "CEO path" means a named person in the network who can ask the company's CEO and has agreed to. Coverage is CEO paths divided by the firm count, times 1,000. Each row then gets one reading:
| Reading | Pattern in the row | What it means | Next move |
|---|---|---|---|
| Covered | Several CEO paths, some companies passing | The network already reaches this cell | Screen and ask through the relationship owners |
| Empty network | Firm count above your cells' median, no CEO path | Supply probably exists and nobody you know reaches it | Recruit one connector |
| Thin market | Firm count below the median, no CEO path | Few companies to find, so recruiting may not repay the effort | Confirm the count before deciding |
| Wrong level | Many contacts, few CEO paths | You know managers, not the people who decide | Ask those contacts who knows their CEO |
| Ownership wall | CEO paths exist but almost nothing passes | The cell's records belong to its clients | Revise the thesis cell, not the network |
Cell D shows why the last reading matters. In this example the billing services hold records that belong mostly to the providers they bill for, and SourceX's screening guidance treats client-owned material as usually not a fit. Six CEO paths do not fix that; the rights checklist before a discovery call is the way to test it company by company.
Method note. Measured: the firm count per cell from SUSB (enterprise size classes of 50 employees and up, matched NAICS code, latest data year) and the network's own relationship records, as of the date you build it; this page's version is dated 2026-10-26. "Large" and "small" mean above or below the median firm count of your own cells, so two analysts reach the same reading. Not measured: demand for any record type, record quality, price, or whether SourceX already knows a company. Readings describe your network's reach, never SourceX's interest.
Where the denominator comes from, and four ways it misleads
SUSB reports firms by enterprise size and industry, with a detailed employment-size file, so a fund can total the firms in size classes of 50 employees and up for each cell's NAICS code. Check where the file's classes break, and if none breaks exactly at 50, write down the boundary you used. Four cautions apply before any gap is taken seriously.
- Firms, not sites. Count enterprises by enterprise size, not establishments. A 400-person company with ten locations is one company to ask, and one CEO.
- March, not peak. SUSB employment is for the week of March 12, while the program reads full-time employees at peak, with contractors excluded. The two can differ for seasonal businesses and for companies that lean on part-time or contract labor.
- One country and an old year. SUSB covers the United States, and its latest listed year is 2022. For other countries use their own statistics and keep the geography rule in the cell definition. 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.
- Activity, not ownership. An industry code describes what a company does, not who owns its records. The ownership-wall reading exists to catch that.
What turns a known company into a covered one
A company counts in the "with a CEO path" column only when four things are written down: the person who will ask, their relationship to the CEO, the CEO as the person who authorizes and signs, and the record owner. It counts as passing only when the published fit facts hold on metadata: 50+ full-time employees at peak, documented operating history, records the company created itself and the rights to license them. Questions stay at record type, rough volume, years, system and administrator; nobody requests files, samples or logins.
Groups need one more line. Under the terms a parent and its subsidiaries are one company, so the USD 50,000 floor applies once, to the group's first qualifying transaction, however many entities license; further qualifying transactions from any entity within 12 months pay 15%, with no floor and no cap. Who gets credit when SourceX once corresponded with the parent, or when a different partner already introduced a sibling business, is not written into the terms. SourceX settles it company by company when it reviews duplicates, so a matrix should flag such rows rather than assume an answer.
Credit rules shape the test too. Each company yields credit for one partner only: the one whose valid introduction landed first inside the 90-day attribution window, confirmed when the company's verified application comes in. Any earlier application, lead, account or email correspondence between SourceX and the company makes it a duplicate, credited to no later partner. The partner who submits must have signed the SourceX Channel Partner Agreement beforehand; one signed under the older name, Channel Partner Non-Circumvention and Non-Disclosure Agreement, still applies as signed.
A two-cell connector test, with the outcomes that change the thesis
Choose the two "empty network" cells with the largest firm counts. For each, recruit one connector: an LP whose own business operates in the industry, a former operator on your advisor bench, a portfolio founder who sells into it. The advisor coverage audit is a good way to find that person.
| Step | Action | Record |
|---|---|---|
| 1 | Brief the connector on the cell and the published fit facts | Date, connector, cell |
| 2 | Ask for up to five CEOs the connector would personally ask | Names offered |
| 3 | Confirm each path with the connector | CEO paths confirmed |
| 4 | Screen on metadata through the connector | Companies passing, and the reason each one fails |
| 5 | The connector asks the CEOs who pass | CEOs who agree to talk |
Write the rules before step 1 and stop after 30 days:
- Keep the cell if at least two companies pass on metadata and one CEO agrees to a conversation.
- Reclassify it as an ownership wall if most named companies hold their clients' records.
- Mark it unreachable for now if the connector cannot name a CEO; the gap is real, but your network cannot close it cheaply.
If both test cells turn out to be ownership walls, the thesis has the wrong industries: the records it needs sit with those industries' clients. Rewrite the cells toward the clients and run the matrix again. The two-list discovery study is the next step if you want to compare your network's coverage with another's.
Where the gap reading could be wrong
- Denominator mismatch. March employment, enterprise definitions and a 2022 data year can put a real company in the wrong size band or miss a new one. Treat counts as orders of magnitude.
- Buyer risk management. NIST's Generative AI Profile (AI 600-1) lists data privacy and intellectual property among the risks unique to or worsened by generative AI. Cells dense in personal data or third-party material may need extra screening or face weaker demand, so a gap can be a cell nobody will license from.
- Program fit. SourceX checks relevance to its current programs in review; criteria differ by program and change over time, so a missing company that fails one program's test could still fit another program later.
- Duplicates. A covered cell can be full of companies SourceX already knows, which no partner can see in advance.
- Connector goodwill. Asking LPs and advisors to open doors spends their credibility. No data on how they respond has been published; record declines in the test.
- The thesis itself. It is our hypothesis, unmeasured by SourceX or anyone else. The matrix is how you find out whether your network can test it at all.
For the venture side of the program, start with the overview for venture platform teams; for finding long-running businesses beside an early-stage portfolio, see how to find mature businesses in a fund's network.
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 connector test turns up a company that passes on metadata and a CEO who wants to talk, get a partner link, test the company again on who qualifies and study the program terms before the connector makes the introduction.
A coverage matrix is network work; deciding whether SourceX itself merits a closer look as an investment is a separate question, and anyone asking it can request investor materials from 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.