Choosing the companies is covered by the ten-company selection matrix, and whether to run a pilot at all by the opportunity cost of another founder program. This page covers what comes between approval and a second cohort.
How should a platform team budget a referral pilot that assumes no revenue?
Approve resources, not returns: the budget lists what the pilot may consume and what each line buys, with no line that pays it back.
What a founder gets is an answer to a narrow question: could records the company created be licensed to SourceX for AI training and evaluation, and does the CEO want to find out? The company applies itself, starting with a short assessment that needs no uploads.
The binding constraint is not analyst time, which a fund can add. It is the attention of the few people who can ask a CEO for a conversation and be heard: a deal partner, an LP principal, a founder whose customer is the target. Each ask draws on their standing with that CEO, so the base case caps asks as well as hours.
The zero-revenue pilot base case (Illustrative, eight companies)
| Line | Kind | Low | High | Borne by | What it buys |
|---|---|---|---|---|---|
| Setup: terms, agreement sample, which entity signs and receives any reward, founder note, operating note | Fixed, once | 6 hours | 10 hours | Head of platform, with counsel | A process someone else can run |
| Screen the company and brief its relationship owner | Per company | 1.5 hours | 3 hours | Platform analyst and lead | A dated fit record |
| CEO conversation, submission and founder questions during review | Per company | 1.5 hours | 3 hours | Platform lead | The CEO's yes, no or later |
| Relationship-owner asks | Per company, capped | 1 ask | 2 asks | Partner, LP or founder | Whether owners would ask again |
| Founder time | Per company, logged outside the fund's budget | 1 hour | 3 hours | The CEO | A usefulness answer |
| Upkeep: recheck the terms version and the partner account | Per quarter | 2 hours | 4 hours | Head of platform | Current rules in every conversation |
| Reward income | Locked | $0 | $0 | Nobody | Nothing in any decision |
For eight companies that comes to 32 to 62 platform hours, 8 to 16 asks and 8 to 24 founder hours for the quarter. Cash is close to nil: partner registration is free, so the likely invoice is counsel's. Every figure is an Illustrative planning input, not a measurement. Approve the range, not a midpoint, so the memo shows where actuals landed.
Upkeep earns its line: the terms' change log lists five versions dated between 12 and 22 October 2026. Income is locked because the terms call portal and calculator estimates projections, not promises, and a reward becomes payable only after the Acquisition Partner's final written acceptance and completed handoff, cleared payment to SourceX, payment of the company, finance approval and the partner's tax and payment details. None of that runs on a platform team's calendar.
What the screen checks before a company earns a budget row
A company gets a row only after a screen built from metadata the fund already holds or can ask the relationship owner about:
- Size. 50 or more full-time employees at peak, contractors excluded.
- Records. Years of operating records the company created itself, in systems someone can name. Material that belongs to its clients is usually not a fit.
- Rights. The company can license the records without breaching client confidentiality, employee privacy or third-party contracts. "Unknown" is an acceptable answer here; SourceX's rights review settles it.
- Location. 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. Licensing talks and the signature are the CEO's call, so the CEO must know about the idea and want the conversation.
Nobody asks for files, samples or logins; the screen records what exists, roughly how much, since when, in which system and who controls it. Funding stage is not a test. Three answers wait for SourceX's review of the company, and the budget assumes none of them: whether a current program needs its records, what a license would cover and whether anything is paid.
Where the platform's work falls across the six referral stages
All of the platform's work sits in the first three published stages; everything that could pay sits in the last three.
| Published stage | Platform work | Who sets the pace | What the memo can count |
|---|---|---|---|
| 1. Register and get a link | Setup; the submitting entity signs the Channel Partner Agreement | The platform | Agreement signed, operating note written |
| 2. Introduce an authorized contact | Screen, brief, ask, CEO conversation, submission | The relationship owner, then the CEO | Owners who asked, willing CEOs, introductions submitted |
| 3. SourceX reviews the opportunity | Answer founder questions; no records handled | SourceX and the company | Verified applications, duplicates reported, reviews open or decided |
| 4. The company completes a qualifying deal | None | The company and SourceX | Plan on none in the pilot quarter; a signature alone triggers nothing |
| 5. Reward becomes payable after the company is paid | None | The Acquisition Partner and SourceX | Dashboard status only |
| 6. Finance approves the reward | Supply the partner's tax and payment details | SourceX finance and the partner | Dashboard status only |
The published stages give no duration for review or deal completion, so a quarter-end memo will show costs in stages 1 to 3 and, at most, dashboard statuses beyond them. That is the expected shape, not a failure; judge the quarter on transitions the platform can see.
Two counting rules keep them honest. A company SourceX already knew, through any earlier lead, account or email correspondence, is a duplicate and is not credited, so it gets its own row (how an existing SourceX relationship affects credit). And an introduction stays open while its 90-day attribution window runs, because credit attaches only when the company submits a verified application.
When is a successful referral pilot ready to expand?
When the good result survives three discounts: it holds among the pilot's weaker companies, enough untouched qualifying companies remain to fill the next cohort, and the process works in someone else's hands. A headline count of willing CEOs settles none of those.
A pilot uses the best-scoring companies a team can find, run by the person most committed to it, and both flatter the result. Economists Omar Al-Ubaydli, John List and Dana Suskind group the threats to scaling a pilot into three: whether the evidence is strong enough to act on, whether the pilot's population resembles the one you will scale to, and whether its situation does. Epidemiologists add regression to the mean: units picked for an extreme baseline score tend to measure closer to the average next time, and the drift can pass for real change. A next cohort drawn from lower-scored companies will resemble the pilot's weaker half, so test expansion there.
Expansion gates and evidence checklist
| Gate | Passes when | Evidence to attach | If it fails |
|---|---|---|---|
| G1. Weaker half | The lower-scored half meets the bar written before the first ask, such as half its CEOs wanting a conversation | Results split by pre-pilot score | Pause and rescore |
| G2. Inventory | Untouched companies passing all four fit facts, with an owner who agreed to ask, at least fill the next cohort | A dated list | Stop and introduce on request |
| G3. Second operator | Someone other than the pilot lead ran two companies from the operating note inside the range | Time log | Pause until tested |
| G4. Relationship owners | No owner reports a strained relationship, and most would ask again | One question on the memo date | Stop if a relationship was harmed |
| G5. Founder usefulness | At least half the CEOs who talked rated it worth their time | Answers gathered by someone who did not pitch | Pause and change the conversation |
| G6. Spend | Hours and asks sit inside the ranges, and the program that gives up hours for expansion is named | Time log | Pause until the hours are found |
| G7. Paperwork | The submitting entity signed the agreement, and the founder note says any reward is paid separately by SourceX, never from the company's payout | The documents | Fix before any new ask |
For G2, count a corporate group once: a parent and its subsidiaries share one reward cap, and whether an earlier contact by one group company affects credit for another is decided by SourceX in duplicate review.
With every gate passed, expand to the smaller of the G2 count and twice the pilot, because each doubling changes who makes the asks and which companies enter. That ceiling is this page's starting default, not a SourceX rule. No gate mentions income.
What goes in a quarter-end review of the referral pilot?
A one-page memo, signed by the two people who approved the base case, that opens with the decision and then shows the evidence in a fixed order. Income gets a row and no weight.
The quarterly decision memo
| Section | What goes in | Rule that keeps it honest |
|---|---|---|
| 1. Decision | Expand, pause or stop, and the date of the next memo | One word first, reasons after |
| 2. Spend | Hours, asks and founder time against each range | Actuals beside low and high |
| 3. Stage transitions | Dated counts at each published stage | Duplicates apart; introductions inside their 90-day window shown as open |
| 4. Founder answers | Was it worth your time? | Asked by someone who did not pitch |
| 5. Gates | G1 to G7 marked pass, fail or untested, with evidence | Untested is never a pass |
| 6. Reward income | Dashboard statuses as counts: Estimated, Eligible, Approved, Paid | Never add up estimates; zero weight |
| 7. Unknowns | What only SourceX's review or the company can settle | No guessed outcomes |
| 8. Sign-off | Head of platform and the approving managing partner | The names that signed the budget |
Section 1 follows the strictest rule that applies:
| Gate results on the memo date | Decision | What follows |
|---|---|---|
| All seven pass | Expand | Next cohort at the size ceiling; the same memo next quarter |
| G2 and G4 pass; another gate fails or is untested | Pause | Open companies keep moving; nobody is added |
| G2 fails | Stop | Introduce a CEO who asks individually; set an inventory trigger |
| G4 fails, or spend tops the range while fewer than half the founders found it useful | Stop | Thank each owner; keep the screening notes |
A pilot that succeeded and should still stop (Illustrative)
Illustrative: an invented fund with invented results.
A growth-stage fund's platform team ran eight companies for one quarter: five portfolio companies, two LP-owned businesses and one held by a co-investor. On the headline count it worked: five of eight CEOs wanted a conversation, and four of those five said it was worth their time. The memo still says stop.
- Decision: stop. No second cohort; the inventory count is rechecked at the annual portfolio review.
- Spend. 51 platform hours against 32 to 62, 13 asks against 8 to 16, 14 founder hours against 8 to 24.
- Stage transitions. Eight owners briefed and seven asked; one portfolio founder declined to raise it with a customer, and nobody went around them. Five CEOs wanted a conversation and four introductions were submitted. SourceX reported one as a duplicate because of earlier email correspondence. Two companies applied and are in review; one introduction is open inside its window.
- Founder answers. Four of five useful, collected by an associate who had not pitched.
- Income. $0. Nothing shows Paid on the dashboard, and estimates are left out.
| Gate | Result | Evidence (Illustrative) |
|---|---|---|
| G1. Weaker half | Fail | Top four by pre-pilot score: 4 of 4 CEOs wanted a conversation; bottom four: 1 of 4 |
| G2. Inventory | Fail | 2 untouched companies pass all four fit facts with a willing owner; the next cohort needs 8 |
| G3. Second operator | Untested | The pilot lead ran all eight |
| G4. Relationship owners | Pass | No strain reported; one LP principal would not ask again this year |
| G5. Founder usefulness | Pass | 4 of 5 |
| G6. Spend | Pass | 51 hours inside 32 to 62 |
| G7. Paperwork | Pass | Agreement signed, founder note in use |
G2 decides it. With two qualifying companies left, a cohort of eight would be filled with companies that fail a fit fact, spending asks on CEOs the program cannot help. G1 says the same thing another way: the result came from the strongest four. Stopping withdraws nobody. The companies in review carry on, and a CEO who asks later is introduced individually by whoever owns the relationship. The revisit trigger is one line: eight untouched companies that pass all four fit facts with a willing owner.
What could make this budget or these gates wrong
- Eight is a small number. A split half is four companies, so G1 is a sanity check, not a statistical test; weak evidence is the first scaling threat Al-Ubaydli, List and Suskind name.
- The weaker-half test can be harsh. A weak bottom half can partly be chance, which is why G1 fails to a pause and a rescore rather than a stop.
- Rules move. Five versions of the terms in one month, and criteria that differ by program, mean every gate result needs a date.
- Demand is checked in review. SourceX assesses relevance to its current data needs, so a pilot can pass every gate and still see its companies declined.
- Goodwill does not show in a time log. No data on how founders or LPs react to these asks has been published; G4 asks them directly.
Method note. The base case, gates, size ceiling and memo are this page's own framework. Program facts come from the SourceX terms (Version 2026-10-22.2, since replaced by version 2026-11-04), the six published referral stages and the screening page; the scaling and regression arguments come from the NBER and International Journal of Epidemiology papers listed below; all were checked 26 October 2026. Hours, asks, thresholds and results are Illustrative. The method needs time logs, founder feedback and dated stage transitions from a permissioned pilot; no such figures are published, nor are SourceX review durations, acceptance rates or payouts. Related reading: hours per qualified introduction compared with other founder services, a company CFO's budget for possible licensing income, the overview for venture platform teams and the guides library.
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.
Once the base case is signed and the gates are written down, register for a partner link, screen each pilot company against who qualifies and read the program terms so the founder note quotes the current version.
Budgeting a referral pilot is platform work; deciding whether SourceX deserves a closer look as an investment is a separate question, and 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.