# Warm introductions vs newsletter opt-ins: a test | SourceX

Canonical: https://refer.sourcex.si/guides/test-warm-introductions-against-newsletter-opt-ins
Last updated: 2026-11-04

A platform team can learn whether a relationship owner's one-to-one ask brings more qualified applications than an opt-in item in its founder newsletter, but only if each company is drawn into one channel before either channel moves. Freeze a list of companies whose CEO reads the newsletter and has a named owner, draw them by lot, send the item only to the opt-in arm's edition, and count verified applications per company drawn on one reading day. No results exist; this is the outline.

The comparison that most teams already hold, newsletter repliers on one side and companies partners chose to ask on the other, cannot answer the question. The two groups were never alike, so the gap between them says as much about who was picked as about the channel.

## Which question this outline answers

It answers one question: for the same kind of company, which first touch leads to more verified applications, a personal ask or a newsletter item the CEO can answer by opting in? Three neighboring questions belong to other designs. Whether a broadcast can test a thesis at all is argued in [why a network experiment beats a portfolio blast](/guides/why-a-network-experiment-is-better-than-a-portfolio-blast). What happens after a CEO has already said yes is tested in [the personal introduction handoff test](/guides/does-a-personal-introduction-improve-handoff-completion). Where the disclosure sits inside an issue is covered by [the member newsletter disclosure guide](/guides/how-to-keep-referral-incentives-transparent-in-a-member-newsletter).

Both channels have a case. A newsletter costs owners nothing, reaches CEOs at a moment they choose, and an opt-in is a strong sign of the CEO's own interest. A personal ask carries context and trust, and it spends a relationship. Write down both predictions before the draw: "the warm arm produces more verified applications" and "the opt-in arm produces as many, with fewer misreadings." Whichever survives the reading day is the one you keep.

## The two-channel randomized test outline

The arms differ in one thing: how the CEO first hears that SourceX exists. Everything after a yes runs the same way.

| Element | Warm arm | Opt-in arm |
|---|---|---|
| First touch | The relationship owner writes or calls the CEO one to one | Edition B of the newsletter carries a short item; edition A, sent to everyone else, does not |
| What the CEO reads | The core lines below, plus why the owner thought of this company | The same core lines, with no company named |
| What counts as interest | The CEO tells the owner they want a conversation | The CEO replies to the named contact in the item |
| After interest | A call with the owner; the company applies itself; a registered partner who has signed the Channel Partner Agreement submits the written introduction | Identical |
| Follow-up | One check-in after ten working days, or none | No reminder in later editions |
| Never sent or asked | Records, samples, headcount figures or a verdict on fit | Identical |

**The core lines, in both arms.** SourceX licenses a company's own business records directly from that company, for use in training and evaluating AI. The company's CEO decides whether to talk. The published screen looks for 50 or more full-time employees at the company's peak, contractors left out, years of records the company created, and the rights to license them. 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. The fund could be paid a reward by SourceX, and section 4 of the terms says it "is paid separately by SourceX and is never deducted from the supplier's agreed compensation." Fit is decided only in SourceX's review, so neither arm says a company qualifies; SourceX's own guidance for portfolio-wide messages is to write "may be a fit; SourceX decides after review."

| Specification | Entry |
|---|---|
| Unit | One company, represented by its CEO |
| Sampling frame | Companies whose CEO already receives the newsletter and has a named relationship owner, that pass the fit facts you can see, with no contact with SourceX you know of; frozen on a date |
| Pre-draw notice | Founders on the list hear in a routine update that the team is comparing two ways of sharing optional programs this quarter, that only dates and replies are recorded, and that anyone can step out before the draw date |
| Draw | By lot within each relationship owner's companies, so every owner has companies in both arms and a gifted writer cannot tilt one side |
| Primary count | Verified applications per company drawn, taken from the stage your partner dashboard shows or from the company itself on day 120 after the send: the 90-day window plus time for review |
| Early read | CEOs who agreed to a conversation with no failed fit fact, per company drawn, on day 45; reported, never decisive |
| Guard counts | Misreadings in the CEO's own words, owner-reported strain, newsletter unsubscribes in the week after edition B, anything sent to the team that came from company records |
| Analysis | Every company stays in the arm it was drawn into, read or unread, replied or silent |

The analysis rule is intention to treat. J-PAL describes it, in paraphrase, as using every participant grouped by original assignment whether or not they received the treatment, which can understate the effect of actually receiving it. Here that cost is worth paying: you cannot tell who read a newsletter, and the team should not try to find out.

## Keeping the newsletter out of the warm arm

A newsletter leaks. Founders forward issues, co-founders share an inbox, and a portfolio chat channel can carry one item to everyone in an hour. Three precautions keep the split honest:

1. **Two editions, not one.** The opt-in arm's CEOs receive edition B; everyone else, the warm arm included, receives edition A, identical except for the item.
2. **No one-to-one ask in the opt-in arm** during the window, unless the CEO opts in first. After an opt-in, the owner's call is the same step the warm arm takes after a yes.
3. **Crossovers are recorded, never corrected.** A warm-arm CEO who says "I saw this in a forward," or an owner who asks an opt-in CEO anyway, is logged. Both companies stay where the draw put them.

If your founders share one busy channel, a company-level draw may be too leaky to read. J-PAL's guidance on spillovers suggests, in paraphrase, randomizing at a higher level, schools rather than students. The equivalent here is drawing whole founder cohorts, such as fund vintages, into each arm, at the cost of needing more companies.

## What counts on the reading day

| Disposition | Warm arm meaning | Opt-in arm meaning | In the denominator? |
|---|---|---|---|
| No first touch | The owner declined to ask | Edition B bounced or was never sent | Yes, reported by reason |
| Silence | No answer by day 45 | No reply by day 45 | Yes |
| CEO declined | Said no to the owner | Replied to say no | Yes |
| Someone other than the CEO | A senior manager is keen; the CEO has not spoken | Same | Yes; routed to the CEO, counted as not yet interested |
| Fit fact fails | Interest, then a published fact fails | Same | Yes |
| Earlier contact with SourceX | Found after the draw | Same | Yes, shown apart; section 3 of the terms makes it a duplicate |
| Applied, not yet verified | On day 120 | On day 120 | Yes |
| Verified application | The primary count | The primary count | Yes |

**Method note.** The outline's parts come from what a channel comparison needs (assignment before exposure, a frozen frame, every assigned company counted, crossovers recorded) and from SourceX's published rules (the CEO decides, one partner per company, dated records decide credit). It measures no demand for records, no record quality and no payout.

Opens and clicks are not on the list. A click proves nothing about what a CEO decided, and section 2 of the terms says "an earlier link click alone does not decide credit." In the opt-in arm, keep the dated reply, the CEO's agreement and the written introduction as separate entries; the reply alone is interest, not an introduction.

## Rules for a small result, written before the send

Illustrative sizing: six owners with eight companies each give 24 companies per arm. Counts that small detect only large differences, and verified applications may run to a handful or none, so the rules are written to say "no verdict" often.

- **Warm wins:** at least four more verified applications than the opt-in arm, at least five in total, and no more misreadings. Keep personal asks as the first touch and use the newsletter for education.
- **Opt-in holds its own:** verified applications equal or higher, misreadings no higher, no owner reporting strain. Make the newsletter the first touch and spend owners' asks on companies that opted in.
- **No verdict:** anything else. Keep the channel with fewer misreadings and say the test did not settle it. The arithmetic behind refusing to call a 4-1 split is set out in [avoiding a winner from five replies](/guides/how-to-avoid-declaring-a-winner-from-five-replies).
- **Veto:** an arm that logs two or more misreadings beyond the other's count is not adopted, whatever its applications.

**Stop early** if an owner reports that an ask damaged a relationship, if a founder objects to being in the comparison after the draw (remove nothing; pause and review), if company records reach anyone on the team, or if more than a quarter of edition B fails to arrive, which makes the opt-in arm a test of deliverability. Nobody is held back for good: after the reading day, owners may ask any opt-in CEO who stayed silent. If a pure no-contact group ever seems necessary, run [the holdout feasibility questions](/guides/when-a-pilot-needs-a-holdout-group) first.

## Reading it without overclaiming

The draw makes the comparison fair for this list, quarter, set of owners and newsletter, and for nothing wider. Reasons CEOs give are descriptive: "I liked hearing it from someone I trust" explains one company, not the gap.

Publish what failed. Both arms' counts appear, zeros included, with the crossover count, the opt-outs before the draw and the version of the core lines used. For how such counts sit next to the program's other measures, see the [conversion benchmark method](/guides/data-referral-conversion-benchmark).

## What neither channel can settle about a company

Neither a warm ask nor an opt-in establishes fit. The company's records are its own; the CEO decides whether to talk; whether the records match current demand, whether customer contracts allow licensing and what any license would cover are settled only in SourceX's review. An introduction the CEO never heard about is, on SourceX's own list, usually a poor fit, so an opt-in from anyone else goes to the CEO before it counts. Nobody on the platform team asks for files, samples or logins in either arm.

## Where this outline can mislead

- **The frame is narrow.** Only companies with both an owner and a newsletter-reading CEO enter. A newsletter also reaches companies no owner knows, and this design leaves that value unmeasured.
- **The notice changes behavior.** Founders told about a comparison may reply differently from founders who were not.
- **Owners differ.** Drawing within owners balances them across arms; it does not make a weak writer strong.
- **One item, one wording.** A different item could do better or worse; the test compares channels as you ran them.
- **The lagging count is thin.** Verified applications at day 120 may be too few for any verdict, which is a finding, not a failure.

## 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 draw, the person who will submit introductions in both arms should [register](/register), check the core lines against [who qualifies](/who-qualifies) and settle duplicate handling under the [program terms](/terms).

Running this test is referral work; whether SourceX is a sound investment is a different decision, and readers weighing it can ask for 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/who-qualifies
- https://www.povertyactionlab.org/resource/elements-randomized-evaluation
- https://www.povertyactionlab.org/resource/data-analysis
- https://refer.sourcex.si/how-it-works
- https://sourcex.si/investors
