The material under test is yours: the one-page explainer, introduction email or short deck you send a CEO before suggesting a conversation with SourceX. The interview tests that material, never the company or the person reading it. Partners' own understanding is tested in the onboarding experiment, and a page's comprehension by role before launch in the landing page test; this design interviews executives after real material has reached them.
Clicks measure attention; misreadings measure harm
An executive who clicks every link can still come away believing something false: that agreeing to a conversation commits the company, that the partner will see its records, that the company has already been approved, or that your reward reduces what the company is paid. Each of those beliefs surfaces later as a damaged relationship or a complaint, and none shows up in a click count.
Health communicators met this problem long ago. AHRQ's teach-back tool asks people to restate in their own words what they need to know or do, and treats a wrong answer as evidence about how clearly something was explained rather than as a test of the listener (paraphrase). Applied to referral content, the executive is never on trial; your page is.
One of these misreadings is more than a clarity issue. Section 4.6 of SourceX's sample partner agreement bars a partner from telling a referred company that interest, an estimate, a signed document or uploaded data makes a purchase or a reward certain (paraphrase). Material that leaves executives believing exactly that has a compliance problem, whatever its open rate.
Decision, versions and the executives who take part
| Element | Specification |
|---|---|
| Decision | Keep your current material or replace it with the revised version |
| Hypothesis | The revised version leaves fewer executives with a critical misreading and does not lower qualified outcomes |
| Version A | Your current material, unchanged |
| Version B | The same facts reordered: what the company decides first, then what SourceX does, then who could be paid, with each term defined in plain words |
| Required in both | The six points in the interview key below, each traceable to a SourceX page |
| Who takes part | An executive at a company you were already introducing, who received the material in the normal course and separately agreed to a twenty-minute interview: the CEO, or a senior executive the CEO asks to take it |
| Assignment | A list of version letters drawn before the first send; each new executive gets the next letter |
| Exclusions | Anyone who saw both versions, and anyone the interviewer works with |
| Interviewer | A colleague with no part in the introduction, reading from a written script |
Both versions must pass the same truthfulness bar before the first send. Each says who could be paid and by whom. Neither implies that the company fits, that SourceX endorses you, that anyone gets preferred access to the company's data, or that anyone but the CEO decides. Neither leads with a reward figure.
The interview is optional and separate from the introduction. Declining it changes nothing about the introduction, and an executive may stop at any point.
The executive comprehension interview
Twenty minutes in three parts, always in this order, so that the open questions come before anything that could teach the answer.
Part 1, free recall (about five minutes). "Without looking at it again, tell me in your own words what the material asked of your company." The interviewer notes which of the six points come up unprompted and asks only neutral follow-ups such as "anything else?". The U.S. Census Bureau describes cognitive interviews in the same spirit: respondents think aloud during or after a questionnaire while interviewers add follow-up probes (paraphrase). The Census guidance also notes that small samples let teams revise wording and retest in rounds; in this test the wording stays frozen until the comparison ends.
Part 2, teach-back (about ten minutes). Six prompts, each answered in the executive's own words.
| # | Prompt | Correct gist | Where the key comes from |
|---|---|---|---|
| 1 | Who at your company would decide whether to talk to SourceX? | The CEO | The program's rule; who-qualifies lists introductions made without the decision maker's knowledge as usually not a fit |
| 2 | Who decides whether your company is a fit? | SourceX, after its own review | Who-qualifies: relevance is checked during review |
| 3 | What would your company need to send right now? | Nothing; the company applies itself and no export or upload is involved | How it works |
| 4 | Could the person who sent this be paid, and by whom? | Possibly, by SourceX, if a qualifying deal completes, and never out of the company's payout | Terms, sections 4 and 5 |
| 5 | Does a meeting or a signature mean SourceX will buy? | No | Terms, section 5; agreement, section 4.6 |
| 6 | Can your company stop, or limit what it offers? | Yes, at any stage | How it works: the company controls whether to proceed and what scope it offers |
Part 3, misreading probes (about five minutes). Four statements, each answered true or false, then "why do you say that?". All four are false.
- "Agreeing to a conversation commits us to license."
- "The person who sent this will see our records."
- "Our company has already been approved."
- "The partner's reward comes out of what we would be paid."
The fourth rests on the terms, which say the reward "is paid separately by SourceX and is never deducted from the supplier's agreed compensation" (section 4).
Scoring. Each teach-back point scores 2 for the correct gist in the executive's own words, 1 for a partial answer or one that needed a neutral follow-up, and 0 for a wrong, missing or "don't know" answer. A critical misreading is a stated false belief about who decides, about approval, about records leaving the company or about the reward reducing the payout, whether it appears in part 2 or part 3. "I don't know" scores 0 but is not a misreading.
Blind coding. The interviewer's notes are stripped of company names and of any wording that reveals the version, then scored by a second person who does not know which version the executive saw. Every fourth interview is scored twice, and the number of agreements is reported with the results.
What the interview never asks. Nothing about the company's records, systems, headcount, customers or plans, and nothing about whether it intends to proceed. Keep the scores; delete the raw notes once coded, unless the executive asked for a copy. If an executive later declines the introduction, record it the way the decline taxonomy describes, without private reasons.
What counts, and what never counts
| Measure | Role | Definition |
|---|---|---|
| No-critical-misreading share | Primary | Executives with no critical misreading, divided by executives interviewed, per version |
| Teach-back score | Secondary | The sum of the six points, from 0 to 12 |
| Unprompted recall | Secondary | How many of the six points came up in part 1 |
| Qualified outcomes | Context only | Companies whose application SourceX verified and credited to your introduction, still open at a read date 120 days after the last interview; reported, never decisive at this sample size |
| Unwanted effects | Guard | Felt pressured, asked not to be contacted again, read the material as SourceX endorsing you, offered company material |
| Opens, clicks, time on page | Never | Attention says nothing about what was understood |
Interview floor, time limit and stop rules
Floor. At least 12 completed interviews per version before the versions are compared. Interviews run for up to six months; if either version is short of 12 by then, report the counts and the most-missed points without a comparison.
Absolute bar, written down before the first interview. Any version in which more than one executive in four shows a critical misreading is revised, whatever the comparison says.
Stop rules.
- An executive believes the company is already approved, or that you will see its records: correct it with that executive before the call ends, pause sends of that version until the cause is fixed, and restart its count.
- An executive says they felt pressured: pause all interviews and review the invitation wording.
- An executive offers documents or data: decline, and log it as a protocol slip.
- SourceX changes a published page behind one of the six keys: re-key the scoring, and never pool interviews from before and after.
- Anyone edits a version mid-test: it becomes a new version and its count starts at zero.
Illustrative arithmetic, not results. Suppose 10 of 12 executives who saw version B and 7 of 12 who saw version A had no critical misreading. If the two versions were equally clear, a gap at least that wide would still appear by chance a little under one time in five (a one-sided exact calculation gives about 0.19). A split of 10 to 5 would turn up by chance about once in 22 tries. Small studies separate versions only when the difference is large; the five-reply guide works through why.
How the interviews would be read
| Pattern | What it means | What you do |
|---|---|---|
| Both versions under the bar and close together | Either version is clear enough | Keep the shorter one, or the current one |
| B has clearly fewer critical misreadings, on the point B reorders | The change explains the gap | Adopt B |
| B has fewer misreadings but more reports of pressure | Clearer, but worse in another way | Keep A, fix B's tone and test again |
| Both versions over the bar | The facts are not landing | Rewrite both around the most-missed point |
| Too few interviews | No comparison is possible | Report counts and missed points; decide on content grounds |
Report everything: executives invited, declined, interviewed and double-scored, the agreement count and each version's misread points. A favorite quote is not a finding.
Ways an interview study misleads
- Polite agreement. Executives often say they understood. Open prompts and the "why" after each probe expose agreement that has no content behind it.
- The interview teaches. Each question hints at an answer, which is why recall comes first and corrections wait until the end.
- Who says yes. Executives who agree to an interview are likely the friendlier ones. Compare versions within that same pool and do not generalize to everyone who received the material.
- Interviewer cues. Blind scoring removes the coder's bias but not the interviewer's tone. A script and a colleague with no stake in the introduction reduce it.
- Understanding is not agreement. An executive who understands every point may still decline. That is a good result for the material: it means the decision was informed.
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 your material clears the misreading bar and a CEO who understood it wants the conversation, register as a partner, test the company once more on the published fit criteria and keep the program terms open while you write the introduction.