"Lead with what's in it for them" is standard sales advice. For a data introduction it carries a specific risk: the company benefit is real but conditional, and a note that opens with it can sound like a promise the program never made. This experiment is built to catch that cost as well as any gain.
The benefit sentence, and what it may never claim
Every claim in the sentence must come from a SourceX page and be true for every CEO who receives it.
| Benefit claim | Published basis | Never stretch it into |
|---|---|---|
| SourceX licenses records from the company that created them and pays the company | SourceX's FAQ and how-it-works pages | Any figure, range or estimate of worth |
| Applying creates no obligation | FAQ: "Applying and the fit check create no obligation." | "No commitment at any stage"; a signed Confirmation does commit the company |
| The company keeps its original records | How-it-works: "You retain ownership of your original licensed data." | "You give nothing up"; selected data is licensed exclusively for AI training |
| An introduction does not reduce what the company is paid | Terms, section 4: the reward is paid separately and never deducted | Any remark on how SourceX prices offers |
The sentence used in both versions (Illustrative): "SourceX licenses operating records directly from the companies that created them, for AI training and evaluation, and pays the company; applying commits you to nothing, and you keep ownership of your original records."
Nothing else about money goes into either version: no amounts or ranges, no promise of speed, no wording that treats a possible license as income the company can plan on, and no hint that the company qualifies or is favored because of who introduced it. SourceX's who-qualifies page tells promoters to say a company may be a fit and that SourceX decides after review.
Question, cohort and the one sentence that moves
Hypothesis, two-sided. Moving the benefit sentence to the first line changes the share of allocated CEOs whose companies submit a verified application, and changes how straight CEOs find the note. The direction is not assumed.
Unit. One CEO per company, contacted by the platform team or a partner in their own right.
Cohort. CEOs contacted in one sending window, drawn from a list written before either version. Leave out companies with any earlier application, relationship or email correspondence with SourceX, which section 3 of the terms treats as duplicates; companies outside the geography SourceX publishes; and CEOs already asked. Allocate alternately down a randomized list, within warm and distant relationships separately.
The blocks. Both versions use the same five blocks, word for word. Only the benefit block changes position.
| Block | Content | Benefit-first order | Benefit-later order |
|---|---|---|---|
| Benefit | The sentence above | 1st | 4th |
| Reason | Why you thought of this company, true for it | 2nd | 1st |
| Decider | The CEO decides; SourceX decides fit after its own review | 3rd | 2nd |
| Disclosure | SourceX may pay the sender a referral reward, which never reduces what the company is paid | 4th | 3rd |
| Ask | Reply yes, no or not now | 5th | 5th |
Sender, subject line, links, send times and follow-up stay identical; the hold-constant worksheet lists every element to freeze. The sender must have signed the Channel Partner Agreement before any introduction is submitted.
Five rival predictions and the count behind each
Write this table into the test log before the first send. HM Treasury's evaluation guidance lists among the typical questions "What measurable outcomes, both intended and unintended, occurred?"; H2 to H4 are the unintended outcomes a benefit lead could produce.
| Prediction | What you would see | The count that shows it | What you would do |
|---|---|---|---|
| H1. Benefit first draws more qualified next steps | More verified applications per allocated CEO, trust no lower | Verified applications credited, divided by CEOs allocated, per version | Adopt benefit-first for this kind of CEO |
| H2. Benefit first draws interest that stops short | More written yeses, the same applications | Yes replies and verified applications, both over CEOs allocated | Keep benefit-later; the extra yeses cost CEOs time |
| H3. Benefit first reads as a pitch | Lower trust scores, more requests to stop | Trust means and stop requests per version | Keep benefit-later and rewrite the sentence |
| H4. Benefit first invites questions nobody can answer yet | More replies asking about amounts or dates | Replies coded as amount or timing questions | Put the decider block next to the benefit and retest |
| H5. Position makes no detectable difference | Counts inside the margin set in advance | All of the above | Keep the version CEOs rated straighter; record no difference |
H5 is the likeliest result in a small network, and it is a finished result, not a failure.
Founder trust and the expectation gap
Replies alone cannot show whether a note set honest expectations. Two later checks can.
Trust, one week after any reply. Someone other than the sender asks every CEO who answered, whether yes, no or not now, to rate two statements from 1 to 5: "The note was straight with me about what SourceX and the sender get" and "The note read like a sales pitch." Answers are reported as means per version, never by name.
The expectation gap, 60 days after an application. For companies that applied, one question with fixed options and no free text: "Has anything you learned since applying differed from what the note led you to expect?" Options: no; yes, about amount; timing; effort for our team; who decides; what is licensed. A benefit lead that wins applications but widens this gap has borrowed trust it will have to repay.
Also count, per version: requests to stop or complaints, replies that misread the note as approval or a promised amount, and any files or records sent to the sender. The target for the last is zero; partners never need company records.
Sample, cutoff, stop rules and the decision rule
Cutoff. Section 2 of the terms credits a company to whichever valid introduction came first within 90 days, and only after that company files a verified application; it adds that "an earlier link click alone does not decide credit". So read the stages on your partner dashboard no earlier than 90 days after the last send, plus a review margin you name in advance.
Floor. Twenty verified applications across both versions, or the end date, whichever comes first. Below the floor, report counts without a winner. The arithmetic for why is in how to avoid declaring a winner from five replies.
Stop or pause when:
- One version draws two requests to stop before the floor. Pause it and review the wording.
- A CEO sends records or files. Return or delete them unopened, say they are not needed, and fix the note.
- The benefit sentence proves untrue for a recipient. Correct it with that CEO and keep them in their assigned version for the analysis.
- SourceX changes the terms version or a page the sentence relies on. Stop and restart with the new wording.
- Never stop because one version is ahead.
Decision rule. Adopt benefit-first only if its verified applications per allocated CEO beat benefit-later's by the margin written in advance, its trust means are no more than half a point lower, and its expectation-gap count is no higher. Otherwise keep benefit-later. A reply-rate gain alone decides nothing; the qualified-reply rubric explains why.
Publishing the version that lost
| Count at the cutoff | Benefit-first | Benefit-later |
|---|---|---|
| CEOs allocated | Not yet collected | Not yet collected |
| Written yes | Not yet collected | Not yet collected |
| Verified applications credited | Not yet collected | Not yet collected |
| Requests to stop or complaints | Not yet collected | Not yet collected |
| Amount or timing questions | Not yet collected | Not yet collected |
| Trust means, both items | Not yet collected | Not yet collected |
| Expectation gap, yes | Not yet collected | Not yet collected |
| Files or records sent | Not yet collected | Not yet collected |
Within 30 days of the cutoff, circulate a readout to everyone who sent or approved the notes: both full texts, the prediction row the counts fit, every stop event, and every allocated CEO in the denominator, including those who never replied. If benefit-first lost, say so in the first line. Nothing about individual CEOs or their companies appears in it, and nothing from it is presented outside the platform as a SourceX result. The founder side of the same question, what a CEO gained whether or not a deal followed, is measured by the founder benefit metric.
Where this design could mislead
- Position is not only position. A first line also sets the note's tone and what the CEO reads next. A result is about this sentence in this place, not about benefits in general.
- CEOs compare notes. Founders in one portfolio may see both versions. Record it where you know; do not drop them.
- The investor's weight. A benefit claim from a backer lands differently from one sent by a peer. A platform team's result may not carry over to an accountant's or an LP's network.
- Applications are not deals. SourceX decides fit in its review, not every qualifying company receives an offer, and section 5 of the terms says a lead, meeting or signed agreement alone earns nothing payable.
- Claims age. The sentence is only as current as the pages behind it; date-stamp it with the terms version.
Methods note. Measured: written replies, verified applications credited by the cutoff, two trust items, one expectation-gap item, stop requests and protocol breaks. Claim sources: SourceX's terms (version 2026-10-22.2, since replaced by version 2026-11-04), FAQ, how-it-works and who-qualifies pages, checked on 2026-10-28. Not measured: company fit, demand for any company's records or any payout. Other designs for network outreach sit in the guides hub.
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 send, register as a partner and sign the agreement, check the benefit sentence against who qualifies and the current program terms, and file the five predictions where nobody can edit them later.
Testing how a referral note opens is a program question. Whether SourceX itself deserves investment is another matter entirely; for that, the SourceX investor overview takes requests for investor materials.
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.