It is for the research or content lead of a partner program, at SourceX or at a network whose members are registered partners. The point is practical: once you know which question stalls, you know which explanation to rewrite and which conversation with a CEO to prepare for.
The six questions under test, quoted from who-qualifies
SourceX's page lists five questions to settle before any promotion, "answered from a conversation with the relationship holder, never from files, samples or logins". The study adds a sixth, taken from the same page's screening profile, as a control: it is a question SourceX says it answers itself.
| # | Question as SourceX publishes it | Who can answer it |
|---|---|---|
| 1 | "What was full-time headcount at its peak, and when?" | The company, through the relationship holder; the page adds that contractors are excluded |
| 2 | "When do the company's own records start, and in which systems?" | The company |
| 3 | "Who created the records, and does any of them belong to clients?" | The company |
| 4 | "Where is the company, and what language are its records in?" | The company; the language condition is explained in SourceX's English-language guide |
| 5 | "Does the CEO know about the idea and want the conversation?" | The CEO |
| 6 (control) | "Relevance to SourceX's current data needs." | SourceX, which "checks this during review" |
The page also says to mark each answer pass, unknown or fail with a date, and that "Unknown is not a fail; it is a question for the relationship holder." The study borrows that spirit: an unknown is a stage to measure, not a mistake to hide.
Four stages, two failure codes
A question is "hard" when a partner cannot get it to a usable state. The study defines four stages a question can reach, plus two codes for answers that go wrong in a different way.
| Stage or code | Definition | What it tells you |
|---|---|---|
| 1. Understood | The partner restates the question in words consistent with the published wording | The question reads clearly |
| 2. Answer known | The partner already knows the answer from the CEO or the company's own statement | No extra conversation needed |
| 3. Knows whom to ask | The partner does not know the answer but can name the person who does and could ask them | One conversation away |
| 4. Stalls | The partner neither knows the answer nor knows whom to ask | The question is hard for this partner |
| Misread | The restatement contradicts the published wording | A comprehension problem, not a knowledge gap |
| Over-claim | On the control question, the partner says they can answer it | Confusion about what SourceX decides |
One rule protects the companies. If the only reason a partner "knows" an answer is something seen through a board or advisory seat, the study codes it as stage 3, with the CEO as the person to ask. Knowledge from that seat can prompt one question to that company's own CEO and nothing more, so it never counts as an answer in hand.
The study's headline measure for each question is the share of partners who stall or misread, shown with its count and denominator. Time to reach a stage is secondary.
The screening question usability study
The design has two parts. Part A measures where each question stalls across many partners; Part B explains why, with a few.
| Element | Part A: structured walk-through | Part B: cognitive interviews |
|---|---|---|
| Who | Every partner in the frame who agrees | Eight to ten partners per round, two rounds, spread across role families |
| Format | Online, about 20 minutes, alone | Video call, about 45 minutes, with a trained interviewer |
| The company | One the partner already knows, chosen privately and never named or described | The same rule |
| For each question | Restate it in your own words; pick the stage reached; time is logged automatically | Think aloud while answering, then scripted probes such as "What does peak mean to you here?" and "Who would you ask, and how?" |
| Order | Questions 1 to 6 shuffled for each partner | Fixed, so interviews compare |
| Between rounds | Not applicable | Revise the study's explanation sheet, never the published questions, and test the revision in round two |
Cognitive interviewing is the established tool for Part B. The National Center for Health Statistics' questionnaire research center uses it as its primary method for evaluating survey questions, examining how people interpret a question, where they answer it incorrectly and whether answers from different groups are comparable (paraphrase). Those are exactly the three things a fit question can get wrong for a partner.
What the study records, and what it refuses to hold
| Recorded | Refused |
|---|---|
| The stage reached, per question | The company's name, sector or location |
| The partner's restatement, with any company detail removed | Headcount figures, record dates, system names |
| Seconds per question | Client names or anything a client owns |
| Role family and months since signing the Channel Partner Agreement | Language shares or other answers the company would give |
| The kind of tie to the chosen company: portfolio, client, former colleague or other | Any record, file, screenshot or sample |
The refusals are not caution for its own sake. Section 4.2 of SourceX's Channel Partner Agreement reads: "This Agreement gives Partner no authority to share a supplier's inventory, records, raw data or personal information, or to access extraction systems." A study that collected answers would be asking partners to do what that sentence withholds. If a partner types a company fact anyway, it is deleted before analysis.
Sample, consent and the partners who drop out
The frame is every partner with a signed Channel Partner Agreement on a fixed date. Everyone is invited to Part A; Part B draws from Part A volunteers, chosen to spread across investors, advisers, consultants and others. The consent screen states the purpose, that no company information is wanted, that results are reported only as group counts, and that the partner may stop at any time.
Every invitation ends in one disposition, and every disposition is reported:
| Disposition | Treatment in the results |
|---|---|
| Completed all six questions | Counted per question |
| Stopped part-way | Counted for the questions answered; the stopping point reported |
| No company in mind | Reported as its own outcome |
| Declined | Counted in the invited total |
| Never responded | Counted in the invited total |
A partner who cannot think of a single company in the network is not noise. Dropping them would make every question look easier and every network look richer than it is.
The report this study will publish
Results: not published. No partner has taken part yet. When the study runs, the report will show, for each question, the count at each stage and code against the number who answered it and the number invited; median seconds per question; restatements coded as misreads, quoted only after company details are removed; and, from Part B, the themes interviewers heard, each with the number of interviews in which it came up. Groups with fewer than 10 answers appear as raw counts only.
| Question | Understood | Answer known | Knows whom to ask | Stalls | Misread | Answered | Invited |
|---|---|---|---|---|---|---|---|
| 1, peak headcount | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| 5, CEO knows and wants it | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected | Not yet collected |
| 6, control (over-claims) | Not applicable | Not applicable | Not applicable | Not applicable | Not yet collected | Not yet collected | Not yet collected |
Differences between role families are descriptive. An adviser who stalls less on question 3 may simply know different companies; the study cannot say that the role causes the difference. The wider practice of reporting every attempt, failures included, follows the research blueprint for company data introductions, and a related design on how company size changes effort is in how company size affects referral workload.
What a hard question would justify, and what it would not
- Question 1 stalls or is misread. Rewrite the partner-facing explanation of "peak" and of leaving contractors out, and tell SourceX if its own wording is the source of the confusion. Never change the size test or invent a workaround.
- Questions 2 to 4 stall. These belong to the company. The fix is a better question for the CEO's team, or pointing the CEO to the company fit checker so the company can screen itself.
- Question 5 stalls. The only remedy is a conversation with the CEO. No study finding makes an introduction acceptable without one.
- The control draws over-claims. Partner materials should say plainly that relevance is SourceX's call, made in review.
The limits are real. Partners may pick an easy company to think about. Restating a question well is not the same as acting on it well. And the results describe one version of SourceX's published questions, on the dates the study ran.
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 every question for a company you know has reached stage 2 or 3 and its CEO wants the conversation, register as a partner, take the questions from who qualifies as published and read the program terms before anyone contacts the company. A study on what makes the introduction itself credible is in what makes a company introduction feel trustworthy, and more designs sit in the guides library.