# Portfolio referral pilot design: pick ten first | SourceX

Canonical: https://refer.sourcex.si/guides/choose-a-ten-company-pilot-before-a-portfolio-wide-launch
Last updated: 2026-11-05

Pick ten companies that let the pilot answer two questions, whether founders find the conversation useful and how many platform hours it costs, rather than ten chosen for the largest imaginable payout. Choose them with a matrix of known fit facts, relationship path, founder bandwidth and dated events, plus two composition rules. Budget hours stage by stage with reward income set to $0, fix the decision date now, and stop if fewer than three of the ten CEOs want a conversation.

A portfolio-wide launch spends platform credibility with every founder at once. Ten companies spend it with ten and tell you whether the program has earned the rest. The design below sticks to the inputs a head of platform actually controls: which companies, whose hours, and which result ends it.

## What the pilot measures, and what it leaves out

The founder benefit is narrow and concrete. A CEO learns whether the company's own operating records could be licensed for AI training and evaluation and, if the CEO wants it, gets an introduction to SourceX, which reviews the company and, if a deal follows, licenses the data and pays the company. Nobody on the platform team handles records. The constraint is just as concrete: hours. Andreessen Horowitz describes its operating team as spanning functions from marketing and talent to legal and policy, all dedicated to helping founders at every stage. Even at that scale, each new service draws on the same operators; on a platform team of three, the draw is felt in a week.

So keep three ledgers apart.

| Ledger | Measured in the pilot? | How |
|---|---|---|
| Founder usefulness | Yes | Three questions to each CEO two weeks after the conversation |
| Platform and network hours | Yes | A time log per company, per stage |
| Reward income | Recorded, never used to decide | Payable only after Acquisition Partner acceptance, cleared Acquisition Partner payment to SourceX, payment of the supplier, finance approval and the partner's tax and payment details |

The base case for the decision is $0 of reward income. The default reward is in the offer block at the end of this page; it should not appear anywhere in the pilot's go or no-go.

## The ten-company pilot selection matrix

Three hard exclusions come first. Any one removes a company from the pilot, however well it would score.

- Fewer than 50 full-time employees at peak, contractors excluded. For a staffing firm, only its own full-time employees count, not the workers it places with clients.
- Main operating records that belong to clients, or that are personal consumer data.
- A company already in contact with SourceX. It is not a new company, so it does not belong in a referral pilot; it carries on with SourceX directly.

Score each remaining company from 0 to 2 on four criteria.

| Criterion | 0 | 1 | 2 |
|---|---|---|---|
| Fit facts known | Two or fewer of size, history, rights and location confirmed | Three confirmed | All four confirmed |
| Relationship path | Nobody can ask the CEO | Reachable through a second person | The relationship owner asks the CEO directly |
| Founder bandwidth | Closing a round, running a sale process or handling a crisis this quarter | Busy but reachable | An ordinary quarter |
| Dated event | None known | An event expected but undated | A documented system migration, archive deletion or retention change with a date |

Two composition rules then shape the ten. No more than two companies may share a main record type, so the pilot learns across record types. At least two must come from outside the funded portfolio, such as LP-owned or co-investor-held companies, so the pilot tests the wider network. Funding status is not one of the program's tests, and a pilot confined to the portfolio cannot show whether the network matters.

Illustrative: a fund's long list of 15 lost three companies to the exclusions (a fintech that peaked at 45 full-time staff, an IT outsourcer whose records belong to clients and a company already talking to SourceX). The other twelve were scored.

| Candidate (Illustrative) | Ring | Main records | Fit | Path | Bandwidth | Event | Total | Pilot |
|---|---|---|---|---|---|---|---|---|
| A. 220-person freight forwarder | Portfolio | Shipment exceptions | 2 | 2 | 2 | 2 | 8 | In |
| B. 140-person home health agency | LP-owned | Visit notes | 2 | 2 | 1 | 2 | 7 | In |
| C. 95-person roofing contractor | Portfolio | Job files | 2 | 2 | 2 | 1 | 7 | In |
| D. 310-person benefits administrator | Co-investor portfolio | Claim files | 2 | 1 | 2 | 2 | 7 | In |
| E. 160-person B2B software company | Portfolio | Support tickets | 2 | 2 | 1 | 1 | 6 | In |
| F. 75-person engineering consultancy | Portfolio | Project files | 2 | 2 | 2 | 0 | 6 | In |
| G. Staffing firm with 70 internal full-time staff | LP-owned | Placement records | 1 | 2 | 2 | 1 | 6 | In |
| H. 120-person dental lab group | Portfolio | Case records | 1 | 2 | 2 | 1 | 6 | In |
| I. 58-person legal operations firm | Portfolio | Matter workflows | 2 | 1 | 1 | 1 | 5 | In |
| J. 180-person B2B software company | Portfolio | Support tickets | 2 | 2 | 1 | 0 | 5 | In |
| K. 130-person B2B software company | Portfolio | Support tickets | 2 | 1 | 1 | 1 | 5 | Alternate: third support-ticket company |
| L. 66-person marketing agency | Portfolio | Campaign files | 1 | 0 | 1 | 1 | 3 | Alternate: no path to the CEO yet |

J and K tie on 5; J goes in because its relationship owner can ask the CEO directly. Three of the ten (B, D and G) sit outside the funded portfolio, which meets the second rule. For the screen that comes before any scoring, the [three-outcome eligibility review](/guides/an-eligibility-review-that-ends-with-a-clear-yes-no-or-later) gives each company a yes, no or later.

## Platform hours at each referral stage, with income at $0

SourceX has published no stage-to-stage conversion rates, so do not budget with guessed ones. Budget an upper bound instead, assuming all ten companies reach every stage the platform touches. The stages follow SourceX's published referral path, from introduction through review.

| Stage | Who works | Planning hours per company | Companies (upper bound) | Hours |
|---|---|---|---|---|
| Screen from data the fund already holds | Platform analyst | 1.0 | 10 | 10 |
| Brief the relationship owner | Platform lead | 0.5 | 10 | 5 |
| Relationship owner asks the CEO | Partner, LP or founder | 0.5 | 10 | 5 |
| CEO conversation about records, metadata only | Platform lead | 1.5 | 10 | 15 |
| Submit the introduction | Platform lead | 0.5 | 10 | 5 |
| Founder questions during SourceX's review | Platform lead | 1.0 | 10 | 10 |
| Setup: read the terms and the Channel Partner Agreement, decide which entity registers and who receives any reward, write the founder disclosure note | Head of platform, with counsel | 6 once | 1 | 6 |
| Upper bound | Everyone above | 5.0 per company plus setup | 10 | 56 |

These hours are Illustrative planning estimates, deliberately generous, not measurements; replace them with your own after the first two companies. A leaner model that counts hours per qualified introduction is in [is data referral work worth a platform team's time](/guides/is-data-referral-work-worth-a-platform-team-s-time). Income in this budget is $0 because the decision date is set in weeks, while a reward can become payable only after a completed deal, Acquisition Partner acceptance and dataset handoff, with payment due within 60 calendar days of the later of those last two.

Track founder hours in their own column. Each CEO conversation also costs the CEO time, perhaps two hours with preparation in this Illustrative plan, and it belongs in the usefulness ledger: two hours a founder rates as useful and two hours a founder resents are opposite results.

## Measure usefulness without asking founders to be polite

Ask three questions two weeks after each CEO conversation, through someone other than the person who pitched it:

1. Did the conversation tell you something about your records that you did not know? Yes or no, then one sentence.
2. Would you want the same conversation offered to another founder you respect? Yes, no or unsure.
3. Was it worth the time it took? Yes or no.

Add two observed measures that do not depend on politeness: whether the company went on to apply, and whether the CEO mentioned it to anyone else. Stages after the application depend on SourceX's review, and the published stages give no review duration, so they may still be open on the decision date. Record them; do not wait for them.

## Expand, pause or stop, on a date set before the first ask

Set the decision date ten weeks after the first relationship-owner ask. The thresholds below are starting defaults chosen by the platform team, not SourceX benchmarks. Write your own before the pilot starts and do not move them afterwards.

| Result on the decision date | Decision | What follows |
|---|---|---|
| Five or more CEOs wanted a conversation, at least four rated it useful, hours within 120% of budget and no relationship owner asked to stop | Expand | Run the next ten to twenty from the same matrix and keep the time log |
| Three or four CEOs wanted a conversation, or several reviews are still open and no founder has complained | Pause | Keep the ten in flight, add nobody and decide again next quarter |
| Fewer than three CEOs wanted a conversation, most rated it not worth the time, hours ran above 150% of budget or any relationship was harmed | Stop | Close politely and keep the screening notes and each company's revisit trigger |

## A pilot that should stop (Illustrative)

Illustrative: an invented fund and invented results.

A seed-to-Series B fund with 38 portfolio companies picks ten: seven portfolio companies and three LP-owned businesses. At week ten:

- Two of the ten CEOs wanted a conversation, and both rated it useful.
- Five declined. Two were closing rounds; three called it "not a priority this year".
- Three fell out late on rights, because their main records turned out to sit in customers' own systems. The fit column had been scored on assumptions.
- The time log shows 71 hours against a budget of 56, or 127% of plan.
- Reward income was $0, as budgeted, and it decided nothing.

The rule says stop, so the team stops. The two companies whose CEOs wanted the conversation stay in SourceX's review; stopping the pilot withdraws nobody. The lesson concerns the portfolio more than the program: too few companies have the size and records yet, and both willing CEOs ran LP-owned businesses. The team writes one revisit trigger, "ten or more portfolio or network companies pass all four fit facts", and checks it at each annual portfolio review.

## What a ten-company pilot cannot tell you

- **Selection flatters.** The ten best-scoring companies overstate how the program would do across the whole portfolio. Read the result as a ceiling.
- **Ten is a small number.** Two willing CEOs against three is noise; the thresholds are a decision rule, not a statistical test.
- **There is no income signal.** A pilot measured in weeks cannot observe rewards, which depend on completed deals and Acquisition Partner acceptance.
- **Criteria move.** SourceX applies program-specific criteria, and a portfolio company that misses this program may suit the next one, so a stop carries a date rather than a verdict for all time.
- **Goodwill is hard to log.** The time log captures hours, not the cost of a founder or LP who felt pushed. Ask each relationship owner directly on the decision date.

**Method note.** The exclusions and the fit column come from the program terms (Version 2026-10-22.2, since replaced by version 2026-11-04, checked 26 October 2026) and SourceX's screening page; the stage list follows SourceX's published referral stages. Bandwidth, dated events, hours and thresholds are the platform team's own planning inputs and are Illustrative here. SourceX has not published conversion rates, review durations or typical payouts, and none is assumed. For where a pilot like this sits in a wider effort, see [running a portfolio-wide data monetization program](/guides/portfolio-wide-data-monetization-program) and [portfolio-wide AI programs in private equity](/guides/portfolio-wide-ai-program-private-equity); the [first-month plan for platform teams](/guides/a-first-month-referral-plan-for-venture-platform-teams) and the [founder email templates](/resources/client-introduction-email-templates-for-venture-platform-teams) help with the first asks. Whether the program deserves platform hours at all is the subject of [the opportunity cost of another founder program](/guides/the-opportunity-cost-of-another-founder-program).

## 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 the ten are chosen and the thresholds are written down, [register for a partner link](/register), recheck each company against [who qualifies](/who-qualifies) and read the [program terms](/terms) before the first relationship owner makes an ask.

Running a referral pilot and evaluating SourceX as a company are separate questions, and a platform head or fund partner interested in the second can request investor materials through the [SourceX investor overview](https://sourcex.si/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.

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/how-it-works
- https://refer.sourcex.si/who-qualifies
- https://a16z.com/about/
- https://sourcex.si/investors