# Follow-ups and referral quality: a study design | SourceX

Canonical: https://refer.sourcex.si/guides/do-more-follow-ups-improve-referral-quality
Last updated: 2026-11-04

Nobody has measured whether extra follow-ups improve the quality of SourceX introductions, and this page reports no result. It lays out a study that could find out: each CEO decides how many reminders they will accept, partners randomize only where a CEO allowed more than one, and quality is judged by applications SourceX verifies that are still moving months later, read next to stop requests and complaints. Until that cohort closes, the evidence says reminders raise replies, which is a different thing.

What a partner can use today is the rule the study is built on: ask the CEO once how they want to be followed up, then never send more than they allowed.

## Reminders raise replies; whether they raise quality is unmeasured

Survey research has studied reminders for decades. The 2023 Cochrane review of ways to raise questionnaire response reports that reminders by text or email were associated with higher response, an odds ratio of 1.60, and telephone reminders for postal questionnaires with an odds ratio of 1.96, with high heterogeneity across several strategies. A questionnaire returned is the whole goal of that research. A referral is different: a reminder that turns silence into a reluctant yes can produce an application that stalls, a CEO who feels pressed, or a company that later withdraws.

The program also limits what any follow-up can achieve. Under section 2 of the terms, credit goes to the first valid introduction of a new company only when that company submits a verified application inside the 90-day attribution window, decided from dated records; a link click alone decides nothing. The application is the company's own act. So the study treats credit as irrelevant and asks a narrower question: among CEOs who permitted it, does a second reminder change the share of introductions that turn into sound, lasting applications?

| Term | How the study uses it |
|---|---|
| Follow-up | Any message the partner starts about the introduction, by any channel, after the CEO's written yes and before the company applies, declines or reaches day 90 |
| Cap | The number of follow-ups the CEO agreed to: none, once or up to twice |
| Allocated CEO | A CEO in the up-to-twice group who was assigned to an arm before any follow-up was sent |
| Quality application | A verified application inside the window whose dashboard stage has moved past verification by day 120, from a company that sent no stop request or complaint |
| Harm signal | A stop request, a complaint to the partner or to SourceX, or a CEO's note that the reminders felt like pressure |

## The cap question, asked once at the CEO's yes

The study begins only after a CEO has agreed in writing to look at SourceX. Reminders to an unanswered first ask stay outside it, because no permission exists yet; the stage rules in [when to stop following up on an inactive referral](/guides/when-to-stop-following-up-on-an-inactive-referral) already limit those to one.

In the message that thanks the CEO for the yes, the partner asks a single question and records the answer word for word:

| Element | Wording or rule (Illustrative) |
|---|---|
| The question | "If the application hasn't come up again on your side, would you like me to check in not at all, once, or up to twice over the next two months?" |
| Default | No answer counts as once, the most cautious setting that still allows a reminder |
| Binding | The cap is a ceiling; the partner may send fewer follow-ups, never more |
| Withdrawal | A stop request at any point lowers the cap to none from that moment |
| Study notice | "I keep the dates of my follow-ups so I can learn how to follow up better. Tell me if you'd rather your company's dates were left out." |
| Never recorded | Why the CEO chose a cap, anything about the company's records, finances or plans |

A CEO who asks to be left out still gets exactly the follow-ups they chose; only their dates stay out of the dataset. If the CEO's request for time names a later date, the [requested-delay follow-up](/guides/a-referral-follow-up-after-a-company-has-asked-for-more-time) governs that row instead, and it leaves the study.

## Randomization lives inside the permission

Three groups form at the cap question. Only one can be randomized.

| Group | Follow-ups sent | Assigned by | What the study can say |
|---|---|---|---|
| Cap none | Zero | The CEO | Descriptive counts only |
| Cap once, or no answer | One, on day 21 if no application | The CEO | Descriptive counts only |
| Cap up to twice, arm A | One, on day 21 if no application | A sequence drawn before enrolment | Compared with arm B |
| Cap up to twice, arm B | One on day 21, a second on day 45 if still no application | A sequence drawn before enrolment | Compared with arm A |

The allocation list is drawn in blocks for each partner before the first CEO enters, so every partner contributes to both arms and a skilled partner cannot tilt one side. Content is held constant: in both arms the day-21 note answers a question the CEO actually asked, as in [the follow-up that adds a useful answer](/guides/the-follow-up-that-adds-a-useful-answer), and arm B's day-45 note is a short close that makes declining easy, in the spirit of [making an ask easy to decline](/guides/how-to-make-a-referral-ask-easy-to-decline). Only the number differs.

Comparisons across groups stay descriptive. CEOs who allow two reminders may simply be keener than those who allow none, so a gap between groups says nothing about what reminders do.

## Quality first, volume beside it, harm as a guard

- **Primary measure:** quality applications per allocated CEO, arm B against arm A.
- **Volume:** verified applications per allocated CEO, reported next to the primary measure and never instead of it.
- **Harm:** stop requests and complaints per allocated CEO, plus any optional one-line note a CEO sends about how the follow-ups felt.
- **Protocol breaches:** any CEO who received more follow-ups than their cap, counted and published.

Partners read their own dashboards, which show each referral's stage and reward status, and they record dates and stages only. No company records, documents or figures enter the dataset, and no partner asks the company anything for the study's sake. Whether a company's records fit a current program is settled in SourceX's own review, independently of how often anyone wrote. Random allocation spreads those decisions across both arms, so a gap in stage movement can be read at the level of the arms, but no single SourceX outcome is ever read as a verdict on one partner's follow-ups.

## Enrolment floor, the day-120 read and the two stop rules

These are study settings, written into the plan before enrolment. Preregistration, as the Center for Open Science describes it, means setting out the research plan ahead of the study and lodging it with a registry, which keeps what was planned apart from what turned up later.

- **Entry window:** two quarters of CEO yeses from consenting partners who have signed the Channel Partner Agreement.
- **Floor:** no comparison is read unless each arm holds at least 40 allocated CEOs; below that, the report gives counts and stops.
- **Read date:** day 120 after the last CEO enters, so every row has passed its 90-day window and the stage check.
- **Breach stop:** one follow-up beyond a cap pauses enrolment until the cause is found and fixed.
- **Harm stop:** if arm B's stop requests exceed arm A's by three or more at a monthly check, or a complaint about follow-up pressure reaches the partner or SourceX, arm B closes and every remaining CEO gets one follow-up at most.

## Every CEO stays in the denominator

The report keeps every CEO who answered the cap question, including those who never replied again, those who declined and those whose companies turned out to be duplicates under section 3. A company SourceX already knew is reported as its own row, because no number of reminders changes that outcome.

| Row in the report | Cap none | Cap once | Arm A | Arm B |
|---|---|---|---|---|
| CEOs entered | Pending | Pending | Pending | Pending |
| Verified applications | Pending | Pending | Pending | Pending |
| Quality applications | Pending | Pending | Pending | Pending |
| Declined by the CEO | Pending | Pending | Pending | Pending |
| No reply by day 90 | Pending | Pending | Pending | Pending |
| Duplicate under section 3 | Pending | Pending | Pending | Pending |
| Stop requests or complaints | Pending | Pending | Pending | Pending |
| Left the study | Pending | Pending | Pending | Pending |

Only the arm A and arm B columns support a causal reading, and only for CEOs who allow two reminders. Released alongside: the registered plan, the cap question as sent, both follow-up texts, the allocation method, breach counts and every zero.

## What each possible result would change

| Result | What partners would do | What it never justifies |
|---|---|---|
| Arm B raises quality applications, harm flat | Keep offering "up to twice" as one of the CEO's options | A second follow-up for CEOs who chose once or none |
| Arm B raises volume, quality flat or lower | Keep one follow-up as the norm; the extra applications did not hold | Counting applications as success |
| Arm B raises harm signals | Drop the second follow-up | Arguing the volume was worth it |
| No clear difference | Prefer the quieter arm, one follow-up | Running the test again until a difference appears |

## Ways this design could still mislead

- **A self-selected stratum.** CEOs who allow two reminders differ from the rest, so a result inside that group may not travel to anyone else.
- **Partners compensating off the record.** A partner in arm A who phones instead of writing breaks the contrast; calls count as follow-ups and are logged.
- **Small numbers.** Even at the floor, a few applications either way can swing a rate; the report gives counts with every rate.
- **Stage lag.** Dashboard stages may update after the read date; the report states the date each stage was read.
- **Moving programs.** SourceX checks relevance to current data needs during review, and programs can change mid-cohort; a dated change is reported, not absorbed.
- **Being watched.** Partners who know their follow-ups are counted may write more carefully in both arms, which narrows any difference.

## 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.

You can adopt the cap question before any study runs: [register](/register), sign the Channel Partner Agreement before submitting introductions, check each company against [who qualifies](/who-qualifies) and the [program terms](/terms), and find related designs in the [guides hub](/guides).

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/how-it-works
- https://refer.sourcex.si/who-qualifies
- https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.MR000008.pub5/full
- https://www.cos.io/initiatives/prereg
