# Unresolved referral reward averages mislead | SourceX

Canonical: https://refer.sourcex.si/guides/why-average-reward-is-misleading-before-most-deals-resolve
Last updated: 2026-11-05

An average reward computed while most introductions are still open describes the few cases that finished first, not the cohort. Declines tend to resolve early and paid rewards late, because payment waits on acceptance, cleared funds, the company's own payment and finance approval, and many in-between amounts are only estimates that the program terms call projections. Report counts by status and the cash actually paid across the whole cohort, show the open cases, and calculate no average until nearly every case has resolved.

The problem is not arithmetic; every number below is correct. The problem is that four correct averages from one cohort run from nothing to more than $60,000, and none of them is the number a fund COO wants.

## Populations, stages and the moment a case counts as resolved

Fix four definitions before anyone computes anything.

| Term | Definition on this page |
|---|---|
| Cohort | Unique companies introduced in one entry window, one row each |
| Resolved | Declined, withdrew, attribution outcome, or every agreed payment to the company made and any reward paid |
| Open | Anything else, including a company paid one installment of two |
| Cutoff | The single date on which every company's status is read |

Reward amounts move through statuses, and only some of them are money. The rewards page lists Estimated, Eligible, Approved, Processing and Paid. Until the company receives its money, a displayed reward is only an estimate built on proposed supplier compensation, and when the company is paid in installments the USD 50,000 element of the reward is paid with the first installment that meets every condition while the 5% element builds up pro rata as each installment is paid. A reward can also shrink after it is paid: should the company's payout later be cut back or reversed, the reward is worked out again in proportion to the reduced payout, and an overpaid reward may be offset against later rewards.

Statisticians have a name for the open rows. In a fixed-time test, NIST's handbook explains, units that have not failed when the test ends are survivors whose failure times beyond the test length are not observed; the data are right censored. An open introduction at a cutoff is the same: its outcome lies to the right of the cutoff, unknown rather than zero.

## One cohort at two cutoffs (Illustrative)

Illustrative and fictional: twelve invented companies whose amounts obey the published formula for a company's first qualifying transaction, the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap, with the USD 100,000 Minimum Transaction applied. Not SourceX data, not a forecast, not anyone's results.

| Company | Status at cutoff 1 | Counted at cutoff 1 | Status at cutoff 2 | Counted at cutoff 2 |
|---|---|---|---|---|
| C1 to C4 | Declined in review | $0 each | Declined | $0 each |
| C5 | Withdrew | $0 | Withdrew | $0 |
| C6 | Company paid $40,000, below the Minimum Transaction; no reward | $0 | Same | $0 |
| C7 | First installment of $120,000 paid; second pending | $56,000 | Both installments paid, $240,000 in total | $62,000 |
| C8 | Estimated on proposed compensation of $400,000 | $70,000 estimate | Closed lower: company paid $280,000 | $64,000 |
| C9 | Estimated on proposed compensation of USD 100,000 | $55,000 estimate | Declined at data fit | $0 |
| C10 | Open, no estimate | Nothing yet | Company paid $640,000; 15% applies | $96,000 |
| C11 | Open, no estimate | Nothing yet | Declined | $0 |
| C12 | Open, no estimate | Nothing yet | Still open | Nothing yet |

At cutoff 1, six companies have resolved (C1 to C6) and $56,000 has been paid. At cutoff 2, eleven have resolved and $222,000 has been paid. Now average it.

| Way of averaging | Cutoff 1 | Cutoff 2 |
|---|---|---|
| Paid amounts divided by resolved companies | $0 (6 companies) | $20,182 (11) |
| Paid amounts divided by all 12, open counted as zero | $4,667 | $18,500 |
| Paid amounts divided by companies that received a payment | $56,000 (1) | $74,000 (3) |
| Paid amounts plus portal estimates, divided by companies with either | $60,333 (3) | $74,000 (3) |

**What went wrong at cutoff 1.** The resolved-only average, $0, is early declines plus one deal below the minimum. The open-as-zero average, $4,667, will rise every month as cases resolve, which looks like progress when nothing has changed. The paid-only average, $56,000, ignores eleven companies. The estimates-included average, $60,333, is pulled up by C8's $70,000 estimate, which later closed at $64,000, and C9's $55,000, which became nothing.

**What even cutoff 2 cannot tell you.** C12 is still open. If it earns nothing, the cohort mean is $18,500; if it closes like C10, $318,000 across 12 companies is $26,500; and because no reward is capped, there is no upper bound at all. And one company, C10, supplies $96,000 of the $222,000 paid: a single large case moves the whole cohort's average.

At cutoff 1 the honest lower bound is $4,667 (every open case earns nothing) and there is no upper bound, because rewards are not capped; even if C7 completes at $62,000 and every other open case closes at the USD 100,000 minimum, the mean is $28,083. A range that open is a finding in itself: the average is not yet knowable.

## Why open cases are not missing at random

Three features of the program make early averages biased in a predictable direction.

1. **Declines come first.** Fit and rights questions are usually settled before any data preparation is authorized, so the earliest resolved rows are disproportionately zeros.
2. **Payments come last.** Five conditions stand between an estimate and cash: the Acquisition Partner's acceptance, cleared funds in SourceX's hands, the company's own compensation paid, finance approval, and the partner's tax and payment details. The 60-calendar-day deadline is counted from completed handoff or from final written acceptance of the complete, validated dataset, whichever is later.
3. **The floor and the uncapped top both distort.** Every first qualifying transaction pays at least USD 50,000, so a small paid cohort clusters near the floor, while no reward is capped, so one large payout such as C10's moves the mean more than typical deal size does.

For what licensing deals pay at the market level, see SourceX's main-site analysis, [what AI data licensing deals typically pay](https://sourcex.si/deals/average-ai-data-licensing-deal-size); this page is about the reward figure inside your own cohort.

## A resolution-first reward report

Replace the average with seven lines, in this order:

1. **Cohort and cutoff.** Entry window, number of unique companies, cutoff date.
2. **Resolution counts.** Declined, withdrew, attribution outcome, below the Minimum Transaction, open (with partly paid cases noted) and fully paid. Example from cutoff 1: 4, 1, 0, 1, 6 of which 1 partly paid, and 0.
3. **Share resolved.** "6 of 12 resolved", in words and counts.
4. **Cash paid to date.** The sum and the number of companies behind it: "$56,000 from 1 company".
5. **Estimates, on their own line.** Labeled projections, never added to cash.
6. **Bounds, if someone insists on a mean.** Lower bound with open cases at zero; a floor scenario with each open case closing at the USD 100,000 minimum less anything already paid; and a plain statement that there is no upper bound, because rewards are not capped.
7. **Distribution, once nearly all have resolved.** Companies at zero, companies whose reward was set by the USD 50,000 floor plus 5%, and companies whose reward was set by 15%, beside any mean.

Do not annualize. A quarter in which one large reward landed says nothing about the next quarter, and multiplying it by four turns one event into a run rate. The [conversion benchmark design](/guides/data-referral-conversion-benchmark) applies the same discipline to rates, and the [expected earnings spreadsheet guide](/guides/referral-expected-earnings-calculator) keeps a partner's own odds apart from completed figures.

**Method note.** The cohort, its statuses and its amounts are invented to show the arithmetic, using the reward formula, Minimum Transaction, installment rule and estimate wording in the program terms and rewards page as of 2026-11-04. The report format is our proposal. Nothing here measures how often introductions convert or how long they take; SourceX has not published those figures.

## Where this reporting approach can still mislead

- **Small cohorts.** With twelve companies, one case changes every line. Report counts with denominators.
- **Reductions after payment.** Paid amounts can be reduced in proportion if a company's payout is reduced or reversed, so "cash paid to date" is a current figure, not a final one.
- **Changing terms.** The terms change by dated version; from version 2026-10-12, for example, the reward calculation applies to all referrals, including earlier ones. Note the version beside each cutoff.
- **Changing demand.** SourceX checks relevance against its current data needs, so a later cohort can resolve differently for reasons outside the fund's control.
- **Thin visibility.** Partners see generic stages or reference codes, not Acquisition Partner prices, so estimates cannot be checked from outside and should be treated as SourceX's indications only.
- **Timing outside anyone's report.** Payment waits on an Acquisition Partner's acceptance and cleared funds. A slow cohort is not necessarily a weak one; the [comparison of payment on collected revenue](/compare/referral-paid-on-collected-revenue) explains why the reward follows cash.

## The companies that belong in a reward cohort

Admit only plausible fits to the cohort: a company whose full-time headcount reached 50+ at its peak (contractors not counted), whose operating history is documented, which holds the rights to license its records, and whose CEO knows about the conversation and authorizes it. 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. Partners gather metadata only, never files or credentials, and what a company's records are worth stays unknown until SourceX reviews them.

Count a parent and its subsidiaries as one company for the once-per-company USD 50,000 floor and the 12-month Repeat Window, which matters when two entities of one group sit in the same cohort; the [renewal guide](/guides/data-referral-renewal-cap) works through installments and parallel licenses under the floor and the Repeat Window. Every introducer must have signed the SourceX Channel Partner Agreement before submitting; agreements signed under its former name, the Channel Partner Non-Circumvention and Non-Disclosure Agreement, stay in force as signed.

## 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 next partner report would show an average, show the resolution counts instead; [register for a partner link](/register), screen new introductions against [who qualifies](/who-qualifies) and keep the [program terms](/terms) version beside each cutoff.

Reporting a referral cohort and assessing SourceX as a possible investment are separate tasks; a reader doing the second should ask for investor materials at 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/terms
- https://refer.sourcex.si/rewards
- https://www.itl.nist.gov/div898/handbook/apr/section1/apr131.htm
- https://sourcex.si/documents/partner/SourceX_Channel_Partner_Agreement_Sample.pdf
- https://refer.sourcex.si/who-qualifies
- https://sourcex.si/investors
