How to test a referral workshop against one-to-one office hours: a matched-cohort event test that pairs similar founder cohorts, flips a coin inside each pair and scores both formats on next steps the CEO asked for, founder trust and host hours

Short answer

Do not compare the two formats on whoever happens to sign up. Pair founder cohorts that resemble each other, flip a coin inside each pair, and give one cohort a group workshop and the other private 20-minute office hours, with the same host, the same facts and the same month. Then compare the cohorts on next steps a CEO asked for in writing, on founder trust from a short anonymous survey, and on host hours per next step. No results from this design exist yet.

Visual briefFull-size diagram
Three rows of paired founder cohorts. In each row a coin between two cohort cards sends one cohort to a group workshop and the other to one-to-one office hours, and every pair feeds the same three measures at the bottom: next steps the CEO asked for, founder trust and host hours.
The design, not a result. Cohort pairs are Illustrative; no founder, partner or SourceX results for this test are published.

It is for a platform or community lead at a venture fund or founder network who already explains data licensing to portfolio CEOs. Whether a session changes anything at all is tested in the now-or-later workshop design; here the session will happen, and the question is which shape earns its hours.

A group room and a private slot fail in different ways

The two formats trade different risks.

FormatHost time for a cohort of 12 foundersWhat it does wellHow it goes wrongWhat the host never does
Group workshopOne live hour plus preparationEvery founder hears every answer, including answers to questions they would not have asked aloudFounders keep company-specific questions to themselves; a room of peers can make interest feel expectedAsk anyone to describe their company's records in front of the others
One-to-one office hoursTwelve 20-minute slots, four live hoursA founder can ask about their own company's situation in privateHours pile up, and the host hears specifics that create a temptation to steerCarry what one founder said into any conversation about another company

The booked office-hour clinic on this site takes the view that private slots suit these questions better than a group. That may hold for your founders, or only for the founders who book; a matched comparison tells you which.

What both formats must say, and what neither may imply

Write one fact sheet before the coin flip and use it in both formats. The formats may differ in shape, never in substance.

Who decides. The company's CEO decides whether it talks to SourceX, and SourceX decides fit after its own review. The fund, the host and any board seat decide nothing for a company.

What the company itself would need. At least 50 full-time employees at its high point, not counting contractors; years of operating records it produced itself; the right to license those records; and a CEO who wants to talk. 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. Whether the records match a current program, whether customer contracts limit their reuse and what a license would cover stay open until SourceX reviews an application.

Who could be paid. If the host is a registered partner, say so in both formats. Section 4 of the terms has SourceX pay any reward separately, never out of the company's compensation, so a founder's company receives the same amount whether or not anyone introduces it.

What nobody gets. No founder, fund or host receives preferential access to a company's data, and nothing in either session suggests SourceX endorses the fund's program; section 4.5 of the sample Channel Partner Agreement bars either side from publicly naming the other as a partner without written authorization.

What never appears. Requests for files, exports, samples or logins (section 4.2 gives a partner no authority to share a supplier's records); a claim that a named company qualifies; any estimate of what a company could be paid, since section 4.6 forbids telling a company that interest or an estimate assures a purchase or reward and section 7.1 bars partners from committing SourceX to payouts.

The matched-cohort event test

ElementSpecification
DecisionWhich format the next cohorts receive by default
HypothesisWritten before the flip and two-sided: the formats differ in CEO-requested next steps per invited company. No direction is assumed
UnitA cohort: founders invited together, such as a fund vintage, an accelerator class or a sector group
MatchingPair cohorts on number of companies, how many have passed 50 full-time employees at peak by public or founder-stated facts, years since founding and how often the host already meets these founders
AssignmentInside each pair, a logged coin flip decides which cohort gets the workshop; the other gets office hours
Held fixedHost, fact sheet, invitation wording apart from the format line, and calendar month
CrossoverA workshop founder who asks for a private conversation gets one; it is logged, and the cohort stays in the workshop arm
Counting windowDays 1 to 60 after the workshop, or after the cohort's last office-hour slot
Read dateDay 150, so a next step on day 60 still has a full 90-day attribution window behind it
AnalysisPair by pair: the difference between the two cohorts in each pair, never founders pooled across pairs

Cohorts, not founders, because founders in one cohort share slides and notes, so splitting a cohort lets each format leak into the other. Pairs, because with few cohorts one unlucky draw can put all the larger, older companies in one arm. Writing in Statistical Science, Imai, King and Nall argue that concerns about pair matching in cluster-randomized experiments lack support and conclude that pairing should be used whenever it is feasible (our paraphrase of the abstract).

Pairing worksheet (Illustrative). Invented cohorts showing the format of the log; no outcomes are implied.

PairCohortCompanies invitedPast 50 full-time at peakCoin result
12022 fintech group114Workshop
12022 logistics group125Office hours
2Spring seed class142Office hours
2Autumn seed class132Workshop
3Growth portfolio, east coast98Workshop
3Growth portfolio, west coast87Office hours

Measures: next steps the CEO asked for, founder trust and host hours

MeasureCounts whenEvidenceNever counts
CEO-requested next step (primary)A founder-CEO asks in writing for an introduction or for the host's partner linkA dated email or messageA booking, a raised hand or a request for the slides
Qualified introductionSourceX verified the company's application, credited it and had not closed it by day 150Dashboard stageThe host's own view of fit
Founder trustThe cohort's average on three survey items, plus the share answering the knowledge item correctlyAn anonymous form run by someone other than the hostAny named answer
Host hoursPreparation, delivery and follow-up time for the cohortA time log kept on the dayEstimates written afterward
Host hours per next stepHost hours divided by CEO-requested next stepsThe two rows aboveA ratio over zero; write "no next step" instead

The trust survey goes to every founder who attended, three days after their session, with four items:

  1. I could ask what I wanted without revealing more about my company than I chose to. (1 to 5)
  2. I felt free to conclude that data licensing is not for my company. (1 to 5)
  3. I trust the fund to keep what I said in this session to itself. (1 to 5)
  4. Who decides whether your company talks to SourceX? (open answer, scored right when it names the CEO)

Report the response rate beside every average. When a founder asks for an introduction, the host confirms in writing what the CEO agreed to share, as the double opt-in introduction describes, and nothing reaches SourceX before then.

Report template, empty until the read date.

Count at day 150Workshop cohortsOffice-hour cohortsPairs where workshop led
Companies invitedNot yet collectedNot yet collectedNot applicable
CEO-requested next stepsNot yet collectedNot yet collectedNot yet collected
Qualified introductionsNot yet collectedNot yet collectedNot yet collected
Founder trust average and response rateNot yet collectedNot yet collectedNot yet collected
Host hours per next stepNot yet collectedNot yet collectedNot yet collected
Crossovers and guard eventsNot yet collectedNot yet collectedNot applicable

Floors, stop rules and what gets published

Read the comparison only after at least four pairs have reached their read date, and show every pair on its own line. Small numbers flatter whoever is ahead: if format made no difference at all, one format would still win all four pairs about one time in eight, and all five pairs about one time in sixteen. A clean sweep of four pairs is a reason to run more pairs, not a verdict. Avoiding a winner from five replies walks through the same trap with email replies.

Stop or pause when any of these happens:

  • an introduction reaches SourceX before its CEO agreed in writing: pause both arms and fix the consent step;
  • anyone offers or receives company files, exports or logins because of a session: stop that format until the wording is fixed;
  • a founder reports feeling pressed, or says they were told their company qualifies: pause that format and rewrite it before the next cohort;
  • SourceX publishes new terms or screening criteria mid-test: teach the current version, record the date and treat later pairs as a new run.

An arm that looks ahead halfway through is never a reason to stop early.

Publish everything: each pair's counts, trust items and response rates, host hours, crossovers, guard events, and the pairs where the format you preferred lost. A test that reports only its wins teaches the next platform team nothing.

Reading the pairs

What the pairs show at day 150Decision for the next cohorts
Office hours lead on next steps and trust, but cost several times the hoursCompare host hours per next step; if office hours still win, offer them, otherwise run the workshop with slots on request
Workshop matches office hours on next steps with fewer hours, trust levelDefault to the workshop and keep private slots available for anyone who asks
One format scores lower on the trust items or the knowledge itemRework that format before comparing anything else; a next step from a confused founder is not one you want
No consistent difference across pairsPick the format your team can deliver accurately every time, and keep logging
Guard events in either armFix the session first; the comparison waits

For what to watch on the day itself, use the event quality funnel.

Where a format comparison can mislead

  • Few cohorts. A platform team may run six or eight cohorts a year; a result from three pairs is a description, not evidence.
  • Matching misses things. Cohorts alike on headcount and age can still differ in sector or in how warmly they regard the fund.
  • The host is part of the treatment. A host who prefers one format may run it better; keep one host per pair and record who it was.
  • Trust is self-reported. Founders may rate kindly a fund that holds their next round; anonymous forms reduce this, nothing removes it.
  • Review sits with SourceX. A qualified introduction is a verified application still open, not an accepted company or a payable reward, and fit depends on SourceX's current programs, which can move during the test.

Methods note. Built from SourceX's terms (version 2026-10-22.2, since replaced by version 2026-11-04), its screening page and the sample Channel Partner Agreement, read on 2026-10-28, plus the pair-matching paper above. It measures founder requests, verified applications, survey answers and host time, not demand, acceptance or payouts, and publishes no figures because none 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.

Before the first coin flip, register and sign the Channel Partner Agreement, build the shared fact sheet from who qualifies and the program terms, and date it so every cohort hears the same version; other test designs are in the guides hub.

Choosing an event format is referral work. Someone weighing SourceX itself as an investment should start instead at the SourceX investor overview, which takes requests for investor materials; SourceX keeps its company and commercial figures for direct conversations with investors.

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 assign formats to whole cohorts instead of to individual founders?

Founders in one cohort talk to each other, share slides and compare notes. If half a cohort got the workshop and half got private slots, each format would leak into the other. Assigning the whole cohort keeps each format intact, and pairing similar cohorts first stops one lucky draw from deciding the result.

Can a founder in a workshop cohort still ask for a private conversation?

Yes, always. Nobody is refused help to protect a test. The request is logged as a crossover, the founder's cohort stays in the workshop arm for the analysis, and the report states how many crossovers each arm had.

Does the host need to be a registered partner to run either format?

No. Explaining the program needs no account. Anyone who later submits an introduction must hold an approved partner account and sign the Channel Partner Agreement first, and each founder hears in both formats that the host could be paid by SourceX if that happens.

Sources

  1. refer.sourcex.si/terms
  2. refer.sourcex.si/who-qualifies
  3. sourcex.si/documents/partner/SourceX_Channel_Partner_Ag…
  4. projecteuclid.org/journals/statistical-science/volume-24/issue…
  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