# Track negative feedback in a referral experiment | SourceX

Canonical: https://refer.sourcex.si/guides/the-referral-experiment-that-tracks-negative-feedback
Last updated: 2026-11-04

In a venture network the most telling negative feedback rarely arrives as a complaint. It reaches a deal partner at a board meeting, hides inside a founder's polite yes to an investor, or shows up as an agreed introduction that never becomes an application. Track it with a two-sided ledger for every company contacted: positive entries on one side, negative entries on the other, and a warning flag on any yes that looks given for the fund's sake. Gather reactions from every channel a founder uses, read each negative as an event, and keep flagged yeses out of any win. No results are published here.

This page is for growth leads on venture platform teams running small tests of how the network talks about SourceX. Complaint coding is one part of trust measurement; this design covers the larger share founders never put in writing.

## Why the worst feedback in a venture network is polite

A fund is not a neutral sender. A founder who depends on its follow-on support, board seat or introductions has reasons to keep every exchange smooth, so a suggestion from an investor can be accepted to avoid friction rather than out of interest. That is a reasoned concern about the setting, not a measured rate, and it shapes the ledger in two ways.

First, some positive outcomes carry a warning. A yes followed by "happy to help the fund" or by weeks of silence is not the same evidence as an eager yes. Second, the founders most uneasy about an approach are the least likely to say so to the person who made it. Feedback has to be gathered from the places it actually surfaces.

[The partner quality scorecard](/guides/the-partner-quality-scorecard-that-includes-company-feedback) rates introducers on company feedback over time, and [the failed introductions field guide](/guides/a-field-guide-to-failed-data-introductions) classifies breakdowns after the fact. This ledger is the outcome record kept during one experiment, built so its decision rule can see what politeness hides.

## The positive-and-negative outcome ledger

**Hypothesis the ledger tests in any experiment it joins.** H0: the version under test produces no more negative entries or warning flags than the current one. H1: it produces more, in which case no gain in yeses can carry it. **Unit:** one company, reached through one CEO someone in the network knows personally, assigned to an arm at random before contact.

Each code is one entry a ledger keeper can log. One company can carry several.

| Code | Kind | Definition | Read |
|---|---|---|---|
| P1 | Positive | The CEO asks in writing to be introduced, or says the company will apply | Day 30 |
| P2 | Positive | The CEO declines in writing in a way that shows the message was understood | Day 30 |
| W1 | Warning on a positive | A yes accompanied by a sign of obligation, such as "anything for the fund" | When logged |
| W2 | Warning on a positive | The CEO agreed, then neither applied nor replied by day 45 | Day 45 |
| W3 | Warning on a positive | After agreeing, the CEO asked someone at the fund whether taking part was expected | When logged |
| N1 | Negative | A complaint about the approach, to anyone at the fund | Within a week |
| N2 | Negative | The CEO says, when asked, that declining did not feel free | Day 21 |
| N3 | Negative | A misreading: the company qualifies, the fund endorses SourceX, someone gets special access to its data, or the fund can decide for it | Within a week |
| N4 | Negative | The CEO learns of an introduction they had not agreed to | Same day |
| N5 | Negative | The relationship owner reports strain with the founder after the ask | Day 30 |
| 0 | Neither | No answer by day 30 | Day 30 |

N3 and N4 follow SourceX's own screening guidance: a line such as "our companies qualify" implies an acceptance nobody has given, and an introduction the decision maker did not know about is usually not a fit.

**Ledger rules.** One dated line per entry, with the experiment arm, the channel it came through and the code. A CEO's own words are kept only with their agreement; otherwise the line holds the code alone. No company records, headcounts or files, no names in any report, and nothing goes to SourceX. Board knowledge never enters: what the fund learned through a board seat or information rights stays where it was learned.

## Where founders voice negatives: the collection map

Fix the channels before launch, and log only what arrives through them.

| Channel | What tends to surface there | Who logs it | Rule |
|---|---|---|---|
| Deal partner or board member | Remarks made in passing at meetings | The partner, within a week | Log the code, never the board discussion around it |
| Platform events and programs | Comments to program staff | Program lead | Same week, code only |
| LP or co-investor who owns the company | Concerns raised by the owner | The colleague who holds that relationship | Only what the owner chose to share |
| Day-21 check-back | Answers to two optional questions | The check-back sender | No reminder, ever |
| Relationship owner at day 30 | Whether anything changed | Ledger keeper | One question per relationship |

Two lines are never crossed. Nobody monitors private founder groups, and nobody asks one founder to report on another.

## A check-back that makes declining easy

At day 21, someone with no stake in the outcome sends every contacted CEO, including those who never replied, one message. It states, only if the fund will stand behind it, that a no to SourceX has no bearing on the fund relationship, and asks two optional questions: "Did you feel free to say no?" and "Did anything about how this reached you feel unwelcome?" A "no" to the first is N2. No reminder follows; a second nudge to collect feedback about pressure is itself pressure.

## Events, not rates, in a small network

Email platforms show how small a negative signal can matter. Google's sender guidance asks bulk senders to keep the spam rate users report below 0.1% and never to reach 0.3% (our paraphrase). Those figures govern bulk mail delivery, not referral messages, but the lesson carries over: when the line that matters sits at a few reports per thousand, a platform team writing to forty founders cannot estimate a rate. It can only notice events.

So every negative entry and every warning flag gets a short written review, and the stop rules below are counts, not percentages. A clean ledger from forty founders is weak evidence that nothing went wrong.

## Stops triggered by ledger entries

| Entry | Action |
|---|---|
| Any N4 surprised CEO | Stop that sender's introductions until a written review closes |
| Any N1 or N2 | Pause the arm it came from and reread its message with someone outside the test |
| Any N3 misreading | Fix the wording in every arm, log the change and restart the count |
| Two warning flags of any kind in one arm | Pause that arm; review whether the ask leaned on the fund relationship |
| Anything about a company reaching someone outside the conversation | Stop the experiment |

A positive entry never lifts a stop.

## The decision rule: flagged yeses never win

Write the margins as numbers before launch, and apply three steps in order.

1. **Harm gate.** Version B can be adopted only if it has no more N entries and no more warning flags than A, and no N4 at all.
2. **Clean outcomes.** Count P1 and P2 entries that carry no warning flag. B must exceed A by the pre-set margin; as an Illustrative form, at least three more per 30 founders.
3. **Introduction floor.** On unflagged P1 alone, B may not trail A by more than the pre-set margin, for example two per 30.

Anything else is inconclusive and the version in use stays. Flagged yeses are reported in full beside the result. They are excluded from winning because a yes the founder did not feel free to refuse measures the fund's leverage, not the message.

## Silence, declines and partial answers

Every company contacted keeps its line. A written decline is P2: the founder understood and chose. Silence is code 0, neither column, and stays in the denominator. A founder who asks to be removed keeps a coded line with no detail. Check-back answers come only from those willing to give them, so report how many answered beside every N2 count.

Random assignment between arms supports reading a difference in entries as caused by the version, for this network and period. Differences between senders, sectors or relationship types are descriptive only.

## One page of the ledger (Illustrative)

Illustrative. Invented entries for generic companies, showing the format only.

| Line | Company code | Arm | Day | Code | Channel | Note |
|---|---|---|---|---|---|---|
| 1 | C-04 | A | 6 | P1 | Asker | CEO asked for the introduction |
| 2 | C-09 | B | 8 | P1 + W1 | Asker | Agreed, adding that he was glad to help the fund |
| 3 | C-11 | B | 9 | P2 | Asker | Declined; systems move this year |
| 4 | C-07 | B | 14 | N3 | Deal partner | Founder thought the fund had pre-cleared the company |
| 5 | C-07 | B | 14 | Action | Ledger keeper | Wording fixed in both arms; count restarted |
| 6 | C-02 | A | 45 | P1 + W2 | Ledger keeper | Agreed on day 5, no application or reply since |

Lines 2 and 6 stay in the report but out of the outcome test. Line 4 changes the message for everyone before line 5 is logged.

## Where the ledger can mislead

- **Unvoiced harm.** Even with five channels, most unease goes unsaid.
- **Uneven reporting.** Partners who pass on remarks diligently make their arms look worse. Report entries by channel.
- **Over-reading warmth.** A W1 flag rests on a judgment about tone. Have a second keeper code flags blind and report agreement.
- **Small numbers.** Most comparisons will be inconclusive; [the zero-transaction learning review](/guides/what-to-do-when-a-pilot-has-no-completed-transactions) shows what can still be learned.
- **Publication pressure.** [The retrospective outline](/guides/how-to-publish-a-pilot-retrospective-investors-can-trust) shows how to publish the negative column and every flag in full.
- **A borrowed benchmark.** Gmail's thresholds describe bulk email; no published threshold exists for referral messages.

**Methods note.** The codes, channels, stops and decision rule were written on 2026-10-28 from SourceX's screening page and program terms, with Google's sender guidance used only to show how small a negative-feedback line can be. Message tests such as [the eligibility messaging test](/guides/how-to-test-eligibility-messaging-without-hiding-restrictions) can carry this ledger as their outcome record. SourceX publishes no results from it.

## What the ledger never decides about a company

The ledger records how founders reacted to being asked; it never judges whether a company fits. Each company's CEO makes the call on talking to SourceX and signs if it goes ahead. SourceX's screen asks for at least 50 full-time employees at the company's peak, with contractors left out, a multi-year history of its own records, rights to license them and relevance to a live program. 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. Whoever submits introductions signs the SourceX Channel Partner Agreement first, and each introduction concerns one company whose CEO agreed to it.

## 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.

Before the first message of any network test, [register](/register), agree the collection channels and the check-back wording, confirm fit questions against [who qualifies](/who-qualifies), and read the [program terms](/terms) that govern every introduction the test produces.

Keeping this ledger is referral work. Evaluating SourceX as an investment is a separate decision; materials for that are requested through the [SourceX investor overview](https://sourcex.si/investors).

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.

## Sources

- https://refer.sourcex.si/who-qualifies
- https://refer.sourcex.si/terms
- https://support.google.com/mail/answer/14229414?hl=en
- https://sourcex.si/investors
