Do specialized networks produce better-qualified referrals? A matched network cohort study design that pairs asks about like companies across specialist and general networks, keeps every declined ask in the count and has no results yet

Short answer

Nobody knows yet, and a raw comparison cannot tell you. Specialist networks may simply ask about larger, older companies in sectors where demand happens to run high. This study design compares like with like: each ask from a specialist network is paired with an ask from a general network about a company in the same size band, record span, record type, geography and quarter, matched before anyone sees an outcome. Every declined and unsuccessful ask stays in the count. No results have been collected or published.

Visual briefFull-size diagram
Two columns of circles, specialist asks on the left and general asks on the right, joined in five pairs under a label reading same size, records, type, place, quarter; three pairs with the same result are gray and two split pairs are blue; three unpaired circles sit in a box reading unmatched, kept, reported apart; a legend reads split pair: counted and same result: adds nothing; a tag reads results: none published.
Illustrative structure only. The pairs, circles and colors show how the design reads results; no asks have been logged.

Two earlier pages approach the same thesis from other sides. The 90-day specialist test briefs two groups from one community and counts authorized introductions per member briefed. The partner quality study follows partner cohorts and reports rates within strata. This design asks the question both leave open: when the companies themselves are alike, does the kind of network still make a difference?

Why a raw comparison cannot answer the question

Specialists know their sector. That is the advantage being tested, and it is also the problem. A network of former operators in specialty distribution will ask mostly about distributors; a regional business club will ask about everything from clinics to contractors. If one sector holds more companies that pass SourceX's screening facts, a raw comparison credits the network for the sector.

Three other differences travel with specialization: company size, the age of a company's records and the depth of the relationship behind the ask. The first two sit in the companies; the third sits in the network, and may be the very route through which specialization works. The design treats them differently for that reason.

Terms the pairs are built on

TermMeaning in this study
Specialized networkA network whose membership rule names one industry or one professional function
General networkA network whose membership is not restricted by industry or function
AskA member asks a company's CEO, directly or through one named person, whether the CEO wants a conversation with SourceX
Pre-ask recordFacts about the company and the relationship, logged within 48 hours of the ask and before any outcome
Outcome ladderL0 CEO declined or no reply in 30 days; L1 CEO agreed; L2 introduction submitted; L3 the company submitted a verified application; L4 the last stage SourceX reports to the partner by day 180
Better qualifiedA higher share of asks reaching L3 within 90 days, among matched pairs
Split pairA matched pair in which exactly one of the two asks reached L3
Unmatched askAn ask with no partner in the other network type; kept and reported apart

L3 is the primary outcome because it is the moment the program's terms credit an introduction: section 2 credits the first valid introduction within the 90-day window "when that company submits a verified application." It is not a licensing outcome. Whether a company's records fit a current program, and whether any deal follows, is decided only in SourceX's review.

The matched network cohort study

ElementFixed before the entry window opensWhy it is set this way
NetworksAt least four specialized and four general networks, each through an organizer who consentsOne network of each kind measures two networks, not two types
MembersOpt in before the window; their asks are logged from day oneNobody joins after a good month
Entry windowEvery ask made in six monthsEnough asks to form pairs
Pre-ask recordCompany code; peak full-time headcount band (50 to 99, 100 to 249, 250 or more, unknown); year records start, in bands; record family; geography group; relationship type; direct or through one personFacts the member already knows, with no files, samples or records
Matching keyExact match on headcount band, record-start band, record family, geography group and entry quarter; one-to-oneThe quarter absorbs any change in SourceX's program criteria
MatcherA person who sees no outcomesOutcomes chosen after matching cannot shape the pairs
Analysis APairs matched on company facts onlyThe whole difference between network types
Analysis BPairs also matched on relationship typeThe difference left once ties of equal depth are compared
Balance checkEach recorded fact compared across the paired groups before outcomes are openedShows whether the pairs are truly alike
ReadingCount split pairs each way; report both counts, the concordant pairs and the difference with an intervalConcordant pairs carry no information about the network
ThresholdNo reading below 20 split pairs in an analysisA handful of pairs cannot carry a claim
Follow-up180 days from each ask, then status frozen; open cases reported as openSlow applications are not scored as failures

The matcher rule follows standard practice in the matching literature. Elizabeth Stuart's review of matching methods says outcome values should not be used in the matching process, and calls diagnosing the quality of the matched samples perhaps the most important step. In this design that means the pairs are fixed, and their balance checked, while every outcome column is still sealed.

Geography is one matching fact, recorded in the program's own words: 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. Asks about companies outside that sentence are logged, coded as outside, and kept out of the pairs.

How declined, missing and unsuccessful asks are handled

The angle that earns this design attention is what it refuses to drop.

  • CEO declines and silence are outcomes at L0, counted in every pair they belong to.
  • Withdrawn asks, where the member stops before the CEO answers, stay in the record and are coded as withdrawn.
  • Duplicates. If SourceX already knew a company, section 3 of the terms treats the introduction as a duplicate and credits no one. Such asks stay in the count with a duplicate code.
  • Overlap. A company asked through both a specialist and a general network leaves the pairs and goes to a separate overlap line. The study states no credit rule; SourceX decides overlapping cases in its own duplicate review. Counting one company across three networks shows how to merge such rows.
  • Members who leave. Their logged asks remain.
  • Unknown facts. "Unknown" is its own band, matched only to unknown, and the share of unknowns is reported.
  • Reason codes are logged for asks that stop short of L3.

Even with matching, this is an observational comparison. The pairs can be alike on every recorded fact and still differ in ways nobody logged, such as how persuasive a member is or how the network briefed its people. Results will be written as associations, never as causes.

How the results will be reported

Nothing is reported now, because nothing has been collected. When every logged ask has reached its 180th day, the report will hold five parts, in this order:

  1. The flow of asks: logged, outside the geography sentence, overlap, unmatched, paired.
  2. The balance table for each analysis, showing the paired groups side by side on every recorded fact.
  3. For analyses A and B, split pairs each way, concordant pairs and the difference with its interval, or the word "inconclusive" below 20 split pairs.
  4. Outcomes for unmatched asks and for the overlap line, kept apart from the comparison.
  5. Every change made to this design after the window opened, with its date.

Networks and companies are never named, and any figure resting on fewer than ten asks reads "under 10".

One pair and one unmatched ask (Illustrative)

Illustrative. Coded companies with invented facts; no asks have been logged and no outcome is implied.

RecordSpecialist askGeneral askUnmatched specialist ask
CompanyCode S-014, specialty pharmacy servicesCode G-031, home care agencyCode S-022, hospital supply distributor
Headcount band100 to 249100 to 249250 or more
Records start2008 to 2015 band2008 to 2015 bandBefore 2008 band
Record familyOperationsOperationsOperations
RelationshipWorked there for six yearsPeer through a business clubAdvisory seat
Furthest rung by day 180L3, verified applicationL1, CEO agreed, no applicationL3, verified application
How it is readSplit pair: one count for the specialist sideSame pairReported in the unmatched table only

Two details matter. The pair is split, so it counts; had both asks stopped at L1, it would add nothing to the reading. And the unmatched ask, however promising, never enters the comparison, because nothing in the general networks looks like it. If unmatched specialist asks pile up, that itself says something: specialists may reach companies general networks never see, a finding this design can report but not test. How a sector specialist finds unusual candidates describes that kind of reach.

Readings fixed before any pair is formed

The readings are set now, before any data, so nobody can choose a story afterwards.

If the results showThe reading
No difference in either analysisSpecialization adds nothing once companies are alike; recruit by access to CEOs, not by sector
A difference in A that shrinks in BThe advantage runs through deeper relationships, which a general network can also build
A difference that holds in BSomething beyond relationship depth, such as knowing where records belong to clients, deserves its own test
A difference only in one sectorThe network effect is a sector effect; report it by sector and stop generalizing
Fewer than 20 split pairsInconclusive; extend the window rather than read the pairs
Many unmatched specialist asksThe match cannot speak for those companies; report their outcomes separately

Where the matched design can mislead

  • Program criteria move. SourceX applies the criteria of each program, and a company that misses one program's bar may suit another later. The entry quarter in the matching key softens this; it cannot remove it.
  • Coarse outcomes. Partners see their own referrals through generic stages or reference codes, under section 4.6 of the sample agreement, so rungs beyond L3 will be thin.
  • Recollected facts. The pre-ask record is what a member knows, not a verified record. Errors spread across both arms but can still blur pairs.
  • Volunteer networks. Networks that agree to be studied may already be the well-run ones.
  • L3 is not supply. A verified application says nothing about licensing, payout or demand, none of which this design measures.
  • Small samples. Eight networks over six months may not reach 20 split pairs, and the design says so in advance rather than lowering the threshold.

Methods note. The design draws on SourceX's program terms (version 2026-10-22.2, since replaced by version 2026-11-04) and who-qualifies page, both read on 2026-10-28, and on Stuart's 2010 review of matching methods. It measures asks and the rungs they reach up to a verified application, plus any stage SourceX reports to partners. It does not measure record quality, demand, licensing outcomes, payouts or member effort. Companies appear only under codes, organizers and members consent before the window opens, and no network or company is named in any output; cells under 10 are shown as "under 10". No asks have been logged and no results exist.

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.

If your network logs an ask whose CEO wants the conversation, the member who will submit it should first register and sign the Channel Partner Agreement, compare the company with who qualifies and work from the terms; more network research designs are in the guides hub.

Running this study is network research; deciding whether SourceX is a company worth investing in is a different exercise, and readers weighing that can request investor materials 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.

  1. Step 1Share your linkSend your personal link to a company you know.
  2. Step 2Company appliesThe company applies itself at /apply.
  3. Step 3Company signs and is paidThe company signs, delivers its data and is paid by SourceX under its agreement.
  4. Step 4You get your rewardthe greater of $50,000 plus 5% and 15% of the actual supplier payout, with no cap. A reward becomes payable after SourceX has paid the referred company under its signed agreement and the payout conditions in the program terms are met, including finance approval and your tax and payment details.

Common questions

Why not just compare each network's application rate?

Because the networks may be asking about different kinds of companies. If specialists mostly know older, larger firms in one sector, a higher rate could reflect the companies rather than the network. Pairing like companies first is what lets the comparison speak to the network.

Does SourceX receive the study data?

No. The study is run by the networks that agree to take part, on coded records of their own members' asks. Nobody collects company records, and nothing about a company passes to SourceX through the study.

What if our network has no general counterpart?

Then join with another organization that runs one, or use this design as a template and wait. A specialist network compared only with itself cannot answer the question this page asks.

Sources

  1. refer.sourcex.si/terms
  2. refer.sourcex.si/who-qualifies
  3. sourcex.si/documents/partner/SourceX_Channel_Partner_Ag…
  4. arxiv.org/pdf/1010.5586
  5. sourcex.si/investors

Free resources

Know a company with valuable proprietary data?

Become a referral partner, get your link and introduce the company’s CEO (in an owner-run company, usually the owner). Referral partners can join from most countries. We can't work with partners in countries or regions under U.S. sanctions or on U.S. restricted-party lists. The companies you introduce should be based in the U.S., Canada and the EU, or in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia with at least 70% of their data in English.

Refer a company →

I own a business

Explore licensing your company's data directly to SourceX for AI training and evaluation. Start a short assessment; no uploads needed.

Start an assessment