It is written for a research director at a venture platform or operating network who wants to know whether introductions leak at the moment of transfer or simply fade with time. Earlier designs answer neighboring questions: how long each stage takes and why stalls happen, in the stall study; which obstacles appear between a CEO's yes and an application, in the friction map; and how advisory firms compare with one another, in the handoff benchmark. This one isolates the handover.
What counts as a handoff, and what counts as lost
For this study a handoff is a change in who must act next. Before a company applies, four holders are possible.
| Holder | Holds the next step when | Typical next action |
|---|---|---|
| Partner | The partner has decided to ask a CEO and has not yet asked, or the CEO has asked the partner to make the introduction | Ask, or submit the introduction |
| CEO | The CEO has been asked and has not answered, or has said yes and not passed the work on | Decide, or delegate |
| Company team | The CEO has named someone to complete the application | Apply through SourceX's form |
| SourceX | The company has applied and verification is pending | Verify the application |
Handing the step on is normal. SourceX's process page says the company applies itself, and section 2.2 of the sample Channel Partner Agreement says "Partner need not join every call, email or negotiation." A partner stepping back is a handoff to be measured, not a failure.
| Term | Definition used in the study |
|---|---|
| Custody change | A snapshot at which the holder differs from the previous snapshot |
| Quiet | Two consecutive snapshots, 28 days, with no dated event from the current holder and no decision recorded |
| Declined | A recorded decision to stop: the CEO says no, the partner decides not to ask, or SourceX treats the company as a duplicate under section 3 of the terms |
| Revived | A dated event from any holder after a quiet spell |
| Lost | Quiet on the read date, never revived |
| Applied | A verified application, the event section 2 ties credit to; the introduction leaves the study |
| Open | Moving on the read date |
A declined introduction is not lost. Someone decided, and the decision is a result. "Lost" is kept for introductions nobody closed.
Why fortnightly snapshots, not a stage clock
A stage clock asks how long each step takes. The question here is where loss happens, which needs a record of who held the step at regular intervals, including the intervals in which nothing moved. Snapshots every two weeks turn the record into panel data: the same introductions observed at fixed times.
Multi-state models were built for that shape. The msm package for R, in paraphrase of its documentation on CRAN, fits continuous-time Markov multi-state models to longitudinal data observed at arbitrary times, panel data among them, with covariates that can be constant or change in steps over time. The primary analysis here stays descriptive; a multi-state model with the holder as the state is a planned sensitivity check, not the headline.
The custody snapshot
Five fields per open introduction, on a fixed fortnightly date, and nothing else.
| Field | Values | Recorded from |
|---|---|---|
| Introduction code | A random code, never a company name | Assigned at entry |
| Holder now | Partner, CEO, company team or SourceX | The partner's log; SourceX's stage record once the company applies |
| Holder since | The date the current holder took over | The same |
| Last dated event from the holder | A date | The partner's log for asks and replies; SourceX's records for application and verification |
| Status | Moving, quiet, declined or applied | Derived by the rules above |
A partner usually cannot see whether a CEO has passed the work to a colleague. The primary analysis therefore merges CEO and company team into one company-side holder, and splits them only where the CEO named a delegate in writing. SourceX's side comes from dated stage records it already keeps. The partner dashboard shows each referral's stage and a reward status (Estimated, Eligible, Approved, Paid), but the study uses the dated stage records, not the statuses. No company records, documents, reasons or message contents enter a snapshot.
The stage-transition attrition table
Two tables carry the results. Both are published empty now, so their shape cannot be adjusted once counts arrive.
Table 1: what happens between consecutive snapshots
| Holder at the earlier snapshot | Same holder, moving | New holder | Quiet | Declined | Applied | Intervals |
|---|---|---|---|---|---|---|
| Partner | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| Company side | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| SourceX | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
Table 2: the handover comparison
| Interval | At risk | Went quiet | Quiet rate | Revived by read date |
|---|---|---|---|---|
| First interval after partner to company side | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| First interval after company side to partner | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| First interval after company side to SourceX | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| No custody change, partner holding | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
| No custody change, company side holding | Not yet counted | Not yet counted | Not yet counted | Not yet counted |
The quiet rate in each row is quiet intervals divided by intervals at risk. "Lost in the handoff" would mean the first three rows run above the matching rows with no custody change.
Sampling frame, consent and the denominator
- Entry. Every ask logged by consenting partners during a two-quarter entry window, from the date the partner records a decision to ask a CEO. Starting before the CEO's yes keeps asks that were never answered, and asks never made, in the count.
- Consent. Partners agree in writing before their logs are used. SourceX contributes dated stage records for linked introductions under its own policies. Companies are not contacted for the study and are never named.
- Follow-up. Each introduction is followed for 180 days, and the read date falls 180 days after the window closes. Section 2's 90-day attribution window is drawn on the timeline; applications more than 90 days after the recorded introduction fall outside it and are counted in their own row.
- Kept in. Declines, duplicates, asks the partner never made, introductions with a missed snapshot (flagged, and tested in a sensitivity run) and introductions from partners who stop logging (censored at their last complete snapshot, in a separate row).
- Dropped after entry. Nothing.
Analysis plan, and what it can and cannot claim
- Count every interval by holder, and by whether custody changed at its start.
- Compute each row's quiet rate with a 95% interval. Rows with fewer than 20 intervals are pooled with a neighboring row, and the pooling is reported.
- Report the share of quiet spells that revive before the read date.
- Report lost introductions per 100 asks on the read date, split by who held the step when each went quiet.
- Rerun with the quiet threshold at one snapshot and at three, and fit the multi-state model as a check.
Every comparison is descriptive. Handovers cluster early in an introduction's life, when interest is least settled, so a higher quiet rate after a handover may reflect timing rather than the transfer. Stratifying by weeks since entry reduces that confound without removing it. Nothing here would show that changing how a handoff is done changes the count; that needs an experiment, such as the choice-first handoff test.
What the report will contain
Both tables in full, zeros included; the number of partners, asks and intervals; the share of snapshots missed; the threshold runs; and the versions of SourceX's terms in force during the window, since section 11 lets SourceX publish new ones. Cells with fewer than five introductions are suppressed, and no partner, firm or company can be identified. Until the read date passes, this page is the plan, and any change to it is dated here. Results already held by SourceX are cited only under the standard in partner results research.
Readings that would overturn the handoff story
- No handover effect. If quiet rates in the first interval after a custody change are no higher than in intervals without one, loss is a matter of elapsed time, and "lost in the handoff" is the wrong frame.
- Most quiet spells revive. Then "lost" overstates attrition, and the read date came too soon.
- Most losses happen while the partner holds the step. Then the leak sits upstream of any handoff, in asks never made.
- The company side cannot be split. Then the CEO-to-team handover stays unmeasured, and the report says so.
Risks a fund should weigh before reading anything into it
- Self-selection. Partners who log carefully may also hand off carefully, which would flatter every row.
- Partial logs. A missed entry makes an active partner look quiet. The missed-snapshot share is reported for that reason.
- Small numbers. Two quarters may yield few handovers per row, and pooled rows hide detail.
- Program changes. A change to the terms or the process mid-window splits the cohort at that date.
- Distance from value. An application is not acceptance, a license or a payment. Section 5 of the terms says "A lead, meeting or signed agreement alone does not earn a payable reward."
None of these is a reason to skip the study. Each is a reason to publish with the denominator showing.
Before any count exists
Name the holder of every open introduction in your own records, today. When you step back, say so in writing and name who holds the next step; the introduction handoff email template does both. Keep dated records, because credit is decided from them and a link click alone decides nothing. And ask only about companies that plausibly fit: a full-time headcount that reached 50 at its peak, contractors not counted, years of records the company created itself and is free to license, and a CEO who knows about the idea. 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.
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.
If your network wants to contribute logs once the entry window opens, register as a partner first, check each company the CEO describes against who qualifies and read the program terms on credit and duplicates; other study designs are collected in the guides hub.
Counting handovers is referral research; deciding whether SourceX is a sound investment is a separate exercise, and anyone weighing it can request investor materials 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.