# Credible data referral introduction examples | SourceX

Canonical: https://refer.sourcex.si/guides/how-to-make-the-introduction-useful-without-overselling-the-dataset
Last updated: 2026-11-04

Describe the records the way a reviewer will read them, not the way a pitch would sell them: record type, system, span of years, a rough count, and who told you and when. Leave value, uniqueness, buyer interest and quality verdicts out, because only SourceX's review can establish them. A line such as "about 38,000 closed tickets since 2018 in one helpdesk, per the head of support on 9 October" gives SourceX something to check, and it gives the founder nothing to be let down by later.

Overselling rarely starts as dishonesty. It starts with enthusiasm, a wish to make the recipient's time feel well spent, and a reward that grows with the company's payout. It costs in two directions at once: the CEO hears promises nobody can keep, and SourceX starts its review on a premise the records may not support.

## What the founder agreed to, and what the introduction has to do

A CEO who says yes to an introduction has agreed to one conversation. They have not asked you to market their company. The message has two jobs: let SourceX decide quickly whether a conversation is worth having, and let the CEO recognize their own company in your words. Everything else is decoration.

The next step being requested is small. The company applies, describes its systems and history, and SourceX takes it from there; credit for the introduction follows when that application is verified. Nothing in your message has to win the deal, because the message cannot.

## Ten fit signals, and the ones a partner can honestly report

SourceX's process page lists ten fit signals it uses to judge a company's records. A partner who has spoken with the owner can describe a few of them at metadata level. Most can only be settled in review, and one is invisible to everyone outside SourceX.

| Fit signal | What a partner can honestly report | What only SourceX's review settles |
|---|---|---|
| Volume | An approximate count with its unit, and who gave it | Whether the count holds after export and de-identification |
| Historical depth | The first year in the current system, and any gap | Whether the earlier years are usable |
| Uniqueness | The company's niche, in plain words | Whether similar records exist elsewhere |
| How detailed the work is | Fields the owner mentioned, such as notes, steps or codes | Whether the detail is consistent across records |
| Domain expertise | Who does the work, such as licensed adjusters or engineers | How far that expertise shows in the records |
| Shows how work ended | Whether records carry a closing state or outcome code | Whether those outcomes are reliable |
| Human feedback | Whether reviews, QA scores or approvals are recorded | Whether that feedback can be used |
| Data structure | The system's name, and whether the owner says it exports | Format and preparation effort |
| Clear ownership | Who created the records, and whether client material is mixed in | Rights, after SourceX's rights review and the company's counsel |
| Acquisition Partner demand | Nothing | Entirely SourceX's assessment |

SourceX labels its own sample data summary "Not a valuation" and says final value depends on actual Acquisition Partner demand and due diligence. If the company's future counterparty will not call its own summary a valuation, an introduction should not either. For the step before writing, when you are still working out which records exist, the [metadata-only opportunity map](/guides/metadata-only-referral-opportunity-map) and the guide to [identifying valuable operational data before a referral](/guides/how-to-identify-valuable-operational-data-before-making-a-referral) cover the questions to ask.

## Six evidence-bounded introduction examples (Illustrative)

Illustrative: invented companies and figures. Each oversold line is a version partners actually reach for; each bounded line says the same thing with its limits attached.

| # | Oversold line | Evidence-bounded line | What bounds it |
|---|---|---|---|
| 1 | "Millions of customer conversations" | "About 38,000 closed tickets in one helpdesk since 2018, according to the head of support on 9 October; I have not seen the system" | Unit, span, source, date and what you did not see |
| 2 | "Decades of operating data" | "Job files in the current field-service system go back to 2011; older years are on paper, the COO says" | First year by system, with the gap named |
| 3 | "Nobody else has data like this" | "They administer claims for regional mutual insurers; whether that is unusual for an AI developer is for SourceX to judge" | A fact about the work, with the verdict left to review |
| 4 | "Clean, AI-ready records" | "Each work order has a closing code and technician notes; the operations lead estimates a fifth have no notes" | Fields named and a known gap included |
| 5 | "They own all of it" | "These are the company's own project files; some include client drawings, and the CEO has not yet asked counsel about those contracts" | Ownership stated with its open question |
| 6 | "AI labs will pay a fortune for this" | Cut the line. If asked: "I have no view on value or demand; SourceX sets any figure after review" | Removed, because no partner evidence supports it |

Three habits run through the column on the right. Every number has a unit and a source. Every gap the owner mentioned stays in. And no line passes a verdict that belongs to someone else.

There is a research template behind this discipline. In "Datasheets for Datasets", Timnit Gebru and coauthors propose that every dataset ship with a datasheet documenting its motivation, composition, collection process, recommended uses and more, so that the people choosing a dataset have what they need to judge it. An introduction is a thin slice of that: a few composition and collection facts, at second hand, before anyone has looked. Overselling is writing that slice as if it were the whole datasheet, with a verdict appended. The live guide to [how AI buyers evaluate datasets](/proof/how-ai-buyers-evaluate-datasets) covers what the full assessment involves.

## One introduction, three possible answers (Illustrative)

Illustrative: a 230-person commercial HVAC services group owned by one of a fund's LPs. Its co-founder and CTO tells an operating partner about "twelve years of service tickets." Three things can happen next.

**Interest, not authority.** The CTO is enthusiastic, but he is not the CEO. The operating partner writes nothing to SourceX. He sends the CTO a short description of the program to share with the CEO, and waits.

**The CEO sponsors it.** The CEO reads a draft, strikes the ticket count ("we haven't checked that"), keeps the rest and approves it in writing. The message that goes out:

> {SourceX contact}, meet {CEO name}, CEO of a 230-person commercial HVAC services group, who has agreed to a first conversation about licensing some of the company's operating records. What I know, from the CEO and CTO this month: service tickets and technician notes in one field-service system since 2014, with a closing code on each job; the records are the company's own, though some customer contracts may limit reuse. What I don't know: the volume after export, whether the records fit a current program, or what they would be worth. I may receive a referral reward from SourceX if a license completes and SourceX pays the company; it would never come out of the company's payout. {CEO name} decides every step from here.

**A voluntary decline.** The CEO says the timing is wrong. Nothing is sent. The operating partner notes the date and the reason the CEO gave, and does not raise it again unless the CEO does.

The approved message is shorter than the CTO's pitch and more useful than it. It says what is known, who said it, what is not known and what the sender gains. The live guide to [introducing without endorsing](/guides/introduction-without-endorsement) covers the neutral wording around it; this page covers the description of the records inside it.

## Whose words go into the introduction

A description of a company's records is company information, and the CEO authorizes the licensing discussion. A co-founder's recollection, a CFO's estimate or a system administrator's count can all be accurate, but each is a personal account until the CEO approves it for the message. The same separation runs on your side. Your interest is a reward that grows with the payout, which is exactly why your adjectives should not set expectations. The fund's interest is a portfolio win. Neither belongs in a description the company has to live with.

When the people around you disagree, the CEO decides which lines travel. When the CEO prefers no figures at all, the introduction goes without them; a record type, a system and a span of years are enough for SourceX to decide whether to talk.

## The approved-lines check before anything is sent

1. **Draft the description from your notes**, each line with its source and date.
2. **Send the draft to the CEO**, not to SourceX, and ask them to strike or correct any line.
3. **Keep the approved version, dated.** Attribution is decided from dated records, including written introduction evidence, so the approved message doubles as your record.
4. **Strip what was not approved**, including figures a colleague gave you and anything learned from board materials.
5. **Attach nothing.** No samples, screenshots or exports: partners never ask for or handle raw records, and no records are shared at SourceX's qualification stage.

The fuller checklist in [preparing a qualified company introduction](/guides/how-to-prepare-a-qualified-company-introduction) covers the rest of the message.

## Fit facts the description should not skip

A precise description of a company that does not fit helps nobody. Before drafting, confirm the published checks as far as you can: a full-time headcount that reached 50 or more at its peak, leaving contractors out; an operating history documented in records the company created; rights to license them; and a CEO who wants the conversation. Records that belong to the company's clients are usually not a fit, which is why example 5 keeps its open question. What stays unknown until SourceX reviews the company: whether the records match current demand, what customer contracts allow and what any license would pay.

## Where bounded introductions fall short

- **They can read as lukewarm.** A CEO may feel undersold by a plain description. No data has been published on whether bounded introductions get faster replies than enthusiastic ones; if you send enough of them, measure it.
- **Second-hand facts can still be wrong.** Attributing a count shows where an error came from. It does not prevent the error.
- **Precision does not cure a failed fit fact.** A beautifully bounded description of a 40-person company is still a description of a company below the size threshold.
- **Demand moves.** An accurate description of records nobody needs today stays accurate, and may matter for a later program.

**Methods note.** The six examples are invented. The ten-signal table maps the signals SourceX publishes on its process page (checked 2026-10-26) to what a partner can see at metadata level; that mapping is our judgment, not SourceX's. Not measured: reply rates, review outcomes or founder reactions for bounded versus oversold introductions, none of which SourceX has published. To test it, log each claim in every introduction you send and compare it later with what the company's application and review actually showed.

## 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.

A description the CEO has approved, with a source and a date on every line, is ready to travel; before it does, [set up your partner account](/register), hold the company up against [the qualification page](/who-qualifies) and read through [the referral terms](/terms). The [overview for venture platform teams](/for/venture-platform-teams) covers the rest of the program from the fund's side.

Readers weighing SourceX as a company to invest in face a different decision from any introduction; investor materials are available on request 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://sourcex.si/how-it-works
- https://refer.sourcex.si/who-qualifies
- https://arxiv.org/pdf/1803.09010
- https://sourcex.si/investors
