Two effects hide inside one change
When fit facts arrive in the first message, CEOs who agree may understand the program better, because they read the conditions before saying yes. A different set of CEOs also agrees, because some read the conditions and stop. A comparison limited to the CEOs who said yes mixes the two, and the second effect can manufacture the first.
Illustrative arithmetic, with invented counts and no real network behind them:
| Reading | Before the switch | After the switch |
|---|---|---|
| CEOs asked | 40 | 40 |
| CEOs who agreed | 20 | 12 |
| Agreed while misreading a fit fact | 8 | 2 |
| Misreading among those who agreed | 8 of 20 | 2 of 12 |
| Agreed with no misreading, per CEO asked | 12 of 40 | 10 of 40 |
Misreading among agreeing CEOs fell from 8 in 20 to 2 in 12, which looks like a clear win. Per CEO asked, correctly informed agreements fell from 12 to 10. If some of the 28 CEOs who stopped after the switch misread a fact that would not have ruled them out, the change cost conversations nobody sees unless someone looks. The application-count version of this trap is worked through in when better screening lowers applications but improves outcomes; this page is about what CEOs understood.
Definitions fixed before the switch date
| Term | Definition used in this analysis |
|---|---|
| Switch date | The single day from which every participating member's first message to a CEO carries the up-front explanation, logged before it arrives |
| Up-front explanation | The fit facts in SourceX's own words, a link to the who-qualifies page and the line "SourceX decides after its own review", placed in the first message rather than after the CEO shows interest |
| CEO asked | A company CEO who received a first message about SourceX from a member inside a period, whatever happened next |
| Self-exclusion | A CEO declines and names a fit fact as the reason |
| Mistaken self-exclusion | A self-exclusion the CEO withdraws on re-check, saying the published wording does not rule the company out after all |
The up-front explanation carries the fit facts as SourceX's who-qualifies page states them: a full-time headcount of 50 or more at the company's peak, with contractors left out; years of records the company created itself; the rights to license them; the CEO's own agreement to the conversation; and the geography sentence, word for word: "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." SourceX describes these criteria as screening guidance that varies with its current data needs, so the message never tells a CEO the company fits. Which questions a member can settle, and which only SourceX can, is sorted in which eligibility questions belong with SourceX.
Frame, periods and consent
Frame. Every CEO a participating member asked about SourceX during the study window, taken from the network's own dated log of asks. The log records the network's own activity (date, member, company reference, stage reached) and nothing from the company's systems. Asks sent to anyone but the CEO are analyzed apart.
Periods. At least three monthly readings before the switch and three after. Cochrane's Effective Practice and Organisation of Care group sets the same floor for the interrupted time series studies its reviews include: a clearly defined point when the change happened and at least three data points on each side (our paraphrase of its study-design guidance). Fewer points cannot separate a change at the switch from a trend already under way.
Comparison group. If some offices switch later than others, the late group's months before its own switch serve as the comparison. A network that switches everyone on one day has none and says so.
Consent. Coding a CEO's written words needs that CEO's agreement. Once a company's outcome is settled, the member who asked sends one short note asking whether the CEO's messages may be coded, without names, for this analysis. CEOs from the months before the switch get the same note, so both periods pass through the same filter. A CEO who says no or never answers stays in the count as a row with a stage and no coded text. Coders record the code, never the fact behind it: a reply that mentions a headcount is coded for whether it reads the contractor rule correctly, and the number is not copied.
Nothing in the frame comes from SourceX's records. Members hold their own asks, the replies they received and the updates SourceX sends about their own introductions, which SourceX's sample partner agreement describes as generic stages or reference codes.
The before-after comprehension analysis
Each CEO asked is read at three moments with one codebook: the first reply to the member's message; any later question sent before the CEO agrees, declines or goes quiet; and an optional closing line in the consent note, "Was anything about which companies SourceX works with different from what you expected?"
| Code | Reading | What it sounds like (Illustrative paraphrase) |
|---|---|---|
| C1 | Headcount read correctly | Counts full-time staff at the peak and leaves contractors out |
| C2 | Records read correctly | Talks about records the company created and still holds |
| C3 | Decision read correctly | Treats the CEO's own yes as the decision |
| C4 | Geography read correctly | Applies the published sentence, English condition included |
| M1 | Contractors counted | Adds contractors or placed workers to reach 50 |
| M2 | Client material claimed | Offers files that belong to its customers as its own to license |
| M3 | Someone else decides | Defers to the fund, the board or the member as the one who says yes |
| M4 | Geography misread | Reads the program as U.S.-only, or skips the English condition |
| M5 | Verdict reading | Says or implies the company has been approved |
| N | No fit content | Nothing in the words touches a fit fact |
Two coders read each consented CEO's words with dates, member names and message versions stripped, so neither knows the period unless the CEO quotes the message (those cases are flagged). A reply can carry several codes. Disagreements go to a third reader, and agreement before that step is reported code by code. Code N never counts as understanding. Geography misreadings get their own follow-up guidance in explaining SourceX's geography without turning companies away.
The re-check. Every self-exclusion, in either period, gets one note, once, from the member who asked: SourceX's sentence for the fact the CEO named, the who-qualifies link and one question, does the reason still stand? The answers are: it stands, I misread it, or no reply. The member never asks for the figure behind the reason and never suggests the company fits; if the CEO now wants a conversation, SourceX still decides fit in its own review.
Measures, written down before the first post-switch message
Every measure shares one denominator: CEOs asked in the period.
| Measure | Numerator | Why it is there |
|---|---|---|
| Informed agreements (primary) | CEOs who agreed with no M code at the first or second reading | The change should raise this, not only the share among those who agreed |
| Misread agreements | CEOs who agreed while carrying any M code | The harm the change is meant to remove |
| Self-exclusions | CEOs who declined naming a fit fact | Expected to rise, and not a failure by itself |
| Mistaken self-exclusions | Self-exclusions withdrawn on re-check | The hidden cost of explaining up front |
| Verdict readings | CEOs coded M5 at any reading | The most damaging misreading, tracked alone |
| Silence | CEOs who never replied | Kept in every denominator |
| Applications | Introductions that reached an application in members' dated records | A downstream check; credit and fit remain SourceX's decisions |
Analysis plan. Fit a segmented regression to the monthly readings of each measure: level and slope before the switch, and the change in each at the switch. Where a comparison group exists, subtract its change over the same months. Show counts beside every rate, with intervals, and pool months into quarters whenever a month has fewer than 20 CEOs asked. No measure is added after post-switch data are seen; any change to the plan is dated and published.
Descriptive or causal. The report may say the change caused a shift only when three conditions hold: three or more readings on each side, a comparison group with no matching jump, and no other change landing in the same month. That last condition is a live risk. SourceX's terms record versions dated 12, 19, 21 and 22 October 2026 in section 11, and its who-qualifies page shows an update on 27 October 2026. A switch made in a month when the program's own pages moved is described, never credited with an effect.
The results table stays empty until both periods close
| Measure, per CEO asked | Three months before | Three months after | Comparison group, same months |
|---|---|---|---|
| Informed agreements | Not yet collected | Not yet collected | Not yet collected |
| Misread agreements | Not yet collected | Not yet collected | Not yet collected |
| Self-exclusions | Not yet collected | Not yet collected | Not yet collected |
| Mistaken self-exclusions | Not yet collected | Not yet collected | Not yet collected |
| Verdict readings | Not yet collected | Not yet collected | Not yet collected |
| Silence | Not yet collected | Not yet collected | Not yet collected |
| Applications | Not yet collected | Not yet collected | Not yet collected |
When the last post-switch month closes and every re-check has had 30 days, the report publishes frame counts, consent rates by period, the coded table, the regression output, coder agreement and each withdrawn self-exclusion, with no detail that could name the company. Counterexamples sit beside the headline: months that moved the wrong way, members whose numbers ran against the network's. The reporting conventions in the research blueprint for company data introductions fit this table.
Readings that would change what the network sends
- Informed agreements fall along with misread agreements. The explanation filters out confusion and good conversations together. Keep the content, shorten the delivery, and test moving one fact, such as the contractor rule, to the second message.
- Mistaken self-exclusions cluster on one code. One fact is being presented badly. Rework how it is introduced, keep SourceX's wording for the fact itself, and run the analysis again.
- Verdict readings survive the switch. Put "SourceX decides after its own review" ahead of the fit facts, not after them.
- Nothing moves beyond the comparison group. For this network, timing makes no difference; keep whichever version costs members less time.
Where a before-and-after reading misleads
- Members learn. Partners write better messages with practice anyway, so a gradual rise can pass for an effect at the switch; the slope before the switch is there to catch it.
- The consent filter. CEOs who let their words be coded may read more carefully than those who refuse. Report consent rates by period and by stage.
- Moving criteria. SourceX applies each program's own criteria, and a company that misses one program may fit another later, so the up-front text carries the date it was checked.
- Small networks. A network asking fewer than 20 CEOs a month pools into quarters and may not reach three readings a side within a year; it publishes its design and counts, and no effect.
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 the switch date is logged and the codebook is fixed, register as a partner, check the up-front text against who qualifies on the day each message goes out, and read the program terms so the first message says nothing SourceX has not published. More designs in this series sit in the guides library.
Running this analysis is a decision about how your network asks; judging SourceX as a company to invest in is a separate one, and investor materials are available on request through the SourceX investor overview.
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.