# Referral network discovery advantage study | SourceX

Canonical: https://refer.sourcex.si/guides/why-some-networks-keep-finding-companies-others-miss
Last updated: 2026-11-05

Nobody has measured it yet, so any claimed discovery advantage is a hypothesis. The likely mechanism is specialist access: a network that knows a workflow and the people who run it hears about established, unglamorous companies a broad network never lists. Test it by giving two networks the same fixed universe, counting companies found by one, by both and, by estimate, by neither, and keeping every rejection and duplicate. Program rules on fit and credit are facts; the advantage stays unproven until a complete cohort is observed.

This page is the study design. No cohort has been observed, and SourceX has not published discovery, rejection or duplicate rates. Every number below is Illustrative and exists only to show the arithmetic.

## Facts, conditional outcomes and the thesis

**Known program facts.** An introduced company needs a peak full-time headcount of 50+ (contractors are excluded), a documented operating history, rights to license and a contact with authority to act for it. Credit is limited to one partner per company, the one behind the earliest valid introduction, and it is fixed when the company's verified application arrives inside the 90-day attribution window. Where SourceX already holds an application, relationship or earlier valid introduction for a company, it reviews the company as a duplicate and credits no later partner (SourceX terms, version 2026-10-22.2, since replaced by version 2026-11-04, checked 2026-10-26).

**Conditional outcomes.** Whether a company is accepted, what it licenses and whether any reward follows depend on SourceX's review, a signed agreement and the payment conditions in the terms.

**The thesis.** Networks built on specialist relationships keep finding established companies that broad networks overlook, because they start from a workflow and its practitioners rather than from a contact list. It is a research question.

| Term | Definition used in the study |
|---|---|
| Universe | The fixed set of companies both networks search, defined before the study starts |
| Discovery method | How a network first listed a company: specialist workflow, broad list, inbound or other |
| Owner access | How close the network is to the company's CEO, on a five-rung scale |
| Mandate fit | Passing the published fit facts on what was known at listing, before any SourceX review |
| Overlooked company | A fit-passing company with a CEO path that only one network listed and SourceX did not already know |
| Complete cohort | Every company either network listed in the window, rejected and duplicated ones included |

## What counts as an overlooked company

Not a company the press ignores, and not one that never raised venture money. Fame and funding say nothing about whether a company keeps years of its own records or employed 50 full-time people at its peak. A definition built on them would reward networks for finding obscure companies, fit or not.

The study uses four conditions, all checkable:

1. **Fit at listing.** The company passes the published fit facts on the information available when it was listed.
2. **A CEO path.** The listing network can reach the CEO, directly or through one named person.
3. **Found by one network only.** The other network did not list it in the same window.
4. **New to SourceX.** SourceX's duplicate check finds no earlier application, lead, account or correspondence.

Condition 4 matters most, and networks cannot see it. A company both networks missed may still be a company SourceX already knows. Overlooked by your peers is not the same as new.

## Specialist discovery versus broad-network discovery

| Dimension | Specialist relationship discovery | Broad-network discovery |
|---|---|---|
| Starting point | A workflow, system or trade the network knows | Contacts filtered by title, size or sector |
| First contact | A practitioner who runs the workflow | An executive the network already knows |
| What it sees early | Record depth and years | Company size and the decision maker |
| Likely blind spot | Whether the practitioner's company has a CEO path | Whether the records are deep, or the company's own |
| Likely failure | A rich workflow with no one who can ask the CEO | A reachable CEO whose records belong to clients |

The last three rows are hypotheses the study tests, not findings. For how advisor networks pass owners to one another, see the [professional referral network guide](/for/professional-referral-network-for-advisors); for building one, the [consultant network guide](/guides/how-to-build-a-referral-network-as-a-consultant).

## The complete-cohort discovery study

### Design

- **Fix the universe first.** For example: every company headquartered in three named metro areas, with 50 to 1,000 employees, in four named industries, drawn from one stated list. Both networks search only this universe.
- **Same window, no sharing.** Both networks search for the same 60 days and exchange no names. Referrals between the two would break the overlap estimate, as the next section explains.
- **List everything considered.** Each network records every company it considered, not only the ones it liked, using the instrument below.
- **Blind matching.** A third person matches the two lists on legal name, registration number and address, without knowing which network listed which company.
- **Duplicate check last.** SourceX's duplicate status is recorded only for companies that go on to be introduced, through the normal route.
- **Contact rules.** No company is contacted for the study. A company is approached only the usual way: its relationship owner asks the CEO, and the CEO decides.
- **Consent.** Network members agree in writing to take part, and the study reports companies by type and size band, never by name.

### Estimating what both networks missed

When two lists sample the same universe independently, their overlap estimates how many companies neither found. Epidemiologists use this two-source capture-recapture method to estimate cases that no register caught. The estimated total equals the product of the two list sizes divided by the number of companies on both.

**Illustrative arithmetic, invented numbers.** The specialist network lists 40 fit-passing companies, the broad network 30, and 12 appear on both.

| Quantity | Calculation | Illustrative value |
|---|---|---|
| Listed by both | Matched names | 12 |
| Specialist only | 40 minus 12 | 28 |
| Broad only | 30 minus 12 | 18 |
| Seen by either | 28 plus 18 plus 12 | 58 |
| Estimated fit-passing companies in the universe | 40 × 30 ÷ 12 | 100 |
| Estimated missed by both | 100 minus 58 | 42 |

The estimate rests on assumptions networks routinely break. A paper on disease monitoring with two dependent sources found that positive dependence still leaves the estimate too low, and negative dependence can push it too high, sometimes severely. Two networks whose members belong to the same trade association are likely to be positively dependent, in which case the 42 would understate what both missed. Referrals are a separate problem: when one source refers cases to the other, standard models can fit the data equally well while giving quite different totals. That is why the design forbids sharing names during the window and records every known referral between the networks. Report the estimate as a range with its assumptions, never as a count.

## Record owner access, operating history and mandate fit

The instrument has one row per listed company. Every field is metadata a relationship owner can supply; nothing comes from a company's records.

| Field | Allowed values | Why it is there |
|---|---|---|
| Discovery method | Specialist workflow, broad list, inbound, other | The comparison itself |
| Date first listed | A date | Separates discovery from later learning |
| Relationship owner | A role, never a name | Who could ask the CEO |
| Owner access rung | 0 no path; 1 knows an employee; 2 knows a functional executive; 3 knows the CEO; 4 the CEO agreed to talk | Access measured the same way for both networks |
| Operating history | Years, with the source (owner, registry, website) | A fit fact, and where the number came from |
| Peak full-time headcount band | Under 50; 50 to 199; 200 to 999; 1,000 or more | A fit fact; contractors and placed workers excluded |
| Record ownership flags | Client-owned, heavy personal data, vendor-hosted, none known | The rights questions a partner can see |
| Mandate fit at listing | Pass, fail, unknown | The control variable |
| Duplicate status | Found by both, known to SourceX, new, not yet known | The overlap and the duplicate count |

For staffing firms, only the firm's own full-time employees count toward 50; placed workers on assignment at clients do not. Record both numbers, because SourceX still considers staffing firms whose records are strong ([staffing firm owners](/for/staffing-firm-owners) covers the detail).

## Count every rejected and duplicated company

A study that reports only successes measures how well each network tells stories. Keep the denominators.

| Where a company leaves the cohort | Example reason | Reported as |
|---|---|---|
| Fit at listing | Peak full-time headcount below 50 | Rejected, fit |
| CEO path | A practitioner knows the records; nobody can ask the CEO | Rejected, access |
| CEO decision | The CEO declines a conversation | Declined by the company |
| Rights | The records belong to the company's clients | Rejected, rights |
| Duplicate inside the study | Both networks listed it | Overlap, kept in both counts |
| Duplicate with SourceX | An earlier application, lead or correspondence | Duplicate, not credited |
| SourceX review | Not a match for a current program | Declined in review |

Report each network's count at every row, including zeros. The [rejection reasons study](/guides/data-referral-rejection-reasons-study) gives a taxonomy for declines in review; this study adds the stages before any introduction exists. When two people claim the same company, the guide to [overlapping introductions](/guides/how-to-handle-an-introduction-that-overlaps-with-an-existing-opportunity) explains what happens.

## Which practices would explain a difference

If one network's overlooked companies outnumber the other's, four practices are candidates, each compared only among fit-passing companies:

1. **Starting from a workflow.** Compare overlooked companies from listings that began with a workflow against listings that began with a contact.
2. **Practitioner first, CEO second.** Compare access rungs at listing and 30 days later.
3. **Early metadata questions.** Compare rights rejections where record ownership was asked before reaching the CEO against those where it was asked after.
4. **Network size.** Divide overlooked companies by members active in the window. A bigger network that finds more companies but not more per active member has more reach, not a discovery advantage.

Stratify by headcount band and industry so that one network's lean toward larger companies does not pass for skill. The result that would falsify the thesis: specialist-only companies fail rights or CEO-path checks more often than broad-only ones, so the edge disappears before any introduction.

## Two illustrative paths and a failure

Illustrative. Generic companies, invented facts, no outcomes implied.

**An owner-led company only the specialist found.** A network of elevator maintenance consultants lists a 140-person founder-owned elevator service company with 18 years of work orders and inspection findings in its own system. The broad network never listed it, because nobody there knows a lift engineer. One consultant has advised the founder-CEO for a decade: access rung 3, fit at listing passes.

**An intermediary path only the broad network found.** A broad advisory network lists a 520-person regional property manager through a wealth adviser who knows its CEO. No specialist network had a reason to look at property management. Access rung 3 from day one; the open question is whether maintenance and lease records belong to the company or to the owners whose buildings it manages, which only SourceX's rights review can settle.

**The failure case.** The specialist network's favorite find, a 300-person claims processor it calls overlooked, turns out to have applied to SourceX directly months earlier. It was overlooked by the broad network, not by SourceX, and no partner is credited. A second find, a back-office services firm, processes claims for insurers; those records belong to its clients. Both stay in the count, and both weaken the thesis.

## Turn the findings into a repeatable network review

Once a study has run, five numbers make a quarterly review any network can repeat.

| Number | Definition | What a change tells you |
|---|---|---|
| Listed | Every company considered this quarter | Effort |
| Fit-passing unique | Fit-passing companies only your network listed | Discovery before duplicates |
| New to SourceX | Fit-passing unique companies minus SourceX duplicates | Discovery that counts |
| CEO path | Listed companies at access rung 3 or 4 | Whether discovery can become an introduction |
| Estimated missed | The two-list estimate against a partner network, with its assumptions | How much is left to find |

Publish the review as a one-page sheet with its date, the universe definition and the method note, linked to this design.

## What would change the recommendation

- If SourceX duplicates dominate specialist-only finds, discovery is not the bottleneck; checking duplicate status earlier is.
- If the two lists are strongly dependent, the missed-company estimate is unreliable, and a third, independent list is needed.
- If broad networks match specialists once fit and size are controlled, the apparent advantage was sector mix, not method.

## Disclosure when results are published

Network members who can earn referral rewards have a material connection to SourceX. The FTC's Endorsement Guides FAQ says a connection that could affect how people weigh an endorsement must be disclosed clearly, and that people who earn commissions from links should disclose the relationship. Any published finding, chart or quote from a participant should say so plainly, and the report should state that SourceX operates the referral program the participants belong to.

## 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.

To test the thesis on your own network, run the two-list design with one partner network, and when a listed company's CEO wants the conversation, [register for a partner link](/register), screen it on [who qualifies](/who-qualifies) and read the [program terms](/terms) before introducing anyone. Every introduction requires the signed SourceX Channel Partner Agreement. More method guides sit in the [guides library](/guides).

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/who-qualifies
- https://pubmed.ncbi.nlm.nih.gov/7888444/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC4036210/
- https://www.ftc.gov/business-guidance/resources/ftcs-endorsement-guides-what-people-are-asking
