# Referral network capacity planning model | SourceX

Canonical: https://refer.sourcex.si/guides/when-a-partner-network-grows-faster-than-review-capacity
Last updated: 2026-11-05

When proposals arrive faster than your team can review them, the backlog grows every month and each new partner makes it worse. Estimate arrivals as enabled partners times the proposals each sends a month, estimate capacity as review hours divided by hours per review, and divide the first by the second. Keep that ratio under about 0.8, our planning margin, by pausing recruitment, improving proposals or adding hours, and place your review before the introduction so a queue never spends a company's 90-day window.

The page is written for a head of channel partnerships whose bench of introducers is growing: advisors in a firm, members of an alliance, partner agencies in a software ecosystem. Its program facts were read on 26 October 2026 in SourceX's terms (Version 2026-10-22.2, since replaced by version 2026-11-04) and on the who-qualifies page. The model, its thresholds and every number in the worked months are this page's own and Illustrative.

## Whose review this is, and whose clock a queue would spend

Review capacity here means your team's hours, not SourceX's. Before a partner contacts a CEO, someone on your side checks the proposed company at metadata level against the fit facts, looks it up in your claim log, confirms that the CEO knows about the idea and answers the partner's questions. SourceX's own review starts later, and its timing is unpublished, so it stays out of the model.

Your screen applies the published fit facts and nothing else: a full-time workforce that reached 50 or more at its peak (contractors do not count), several years of records the company produced itself, the rights to license them, and location. 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. Every answer comes from a conversation with the partner, never from files, samples or logins.

Where you put that review decides what a queue costs. The terms credit whichever valid introduction of a new company lands first in the 90-day attribution window, from the moment the company submits a verified application, and they do not lengthen that window because contact protection is still running. A queue before the introduction costs freshness and the CEO's patience. A queue after it costs days a company cannot get back.

Two terms keep the arithmetic honest. Activation is a partner's first proposal that reaches review; [the activation ladder](/questions/how-to-qualify-referral-partners) counts it. Referral quality is the share of proposals that pass your fit screen, which decides how many review hours turn into companies worth introducing.

## The arrival-rate and service-capacity model

The model has two sides. Arrivals grow with the bench: every partner you enable adds a trickle of proposals. Capacity grows only when you add hours or make each review shorter. When arrivals pass capacity, the difference piles up each month, and a growing bench makes the pile grow faster each month.

John Little's 1961 proof in Operations Research gives the steady-state link between a line's length and its wait: the average number of units in a system equals the arrival rate times the average time each spends there, L = λW. It holds only under stable conditions, which a growing backlog is not, so the model uses it for steady months and a plain count for growing ones. Choosing which company takes a scarce slot is a separate problem, covered in [prioritizing introductions when review capacity is limited](/guides/how-to-prioritize-introductions-when-review-capacity-is-limited), and sizing one person's caseload is covered in [how many active referrals one operator can manage](/guides/how-many-active-referrals-can-one-operator-manage).

| Input or output | Symbol | How to measure it | Illustrative value |
|---|---|---|---|
| Enabled partners | N | Signed agreement and a partner link, counted on the first of the month | 20, rising by 10 a month |
| Proposals per enabled partner per month | p | Share of enabled partners who propose in a month times proposals each, from dated records | 0.4 (a quarter propose, 1.6 each) |
| Arrivals | λ = N × p | Proposals reaching your review in the month | 8 in month 1 |
| Hours per review | h | Fit screen, claim-log check, CEO-consent confirmation and partner questions, timed on recent cases | 1.5 |
| Review hours per month | H | Hours left after general partner support | 30 |
| Capacity | μ = H ÷ h | Reviews your team can finish in a month | 20 |
| Utilization | ρ = λ ÷ μ | Arrivals as a share of capacity | 0.4 in month 1 |
| Backlog | B | Last month's backlog plus λ minus μ, never below zero | 0 until month 5 |
| Wait for a new proposal | B ÷ μ | Months of work already in line, converted to weeks | None until month 5 |
| Wait in a steady month | W = L ÷ λ | Average proposals in review divided by arrivals, from Little's formula | 10 in review and 20 arriving gives half a month |

## Eight months against a fixed team (Illustrative)

Illustrative: an invented network that recruits ten partners a month and never adds review hours. Arrivals cross the 0.8 line in month 3 and capacity in month 4. From month 5 the backlog grows faster each month, because each month adds ten more partners' worth of arrivals on top of the last.

Read month 8 twice. A new proposal now waits about nine weeks before anyone screens it. If your team's review sat after the introduction instead of before it, that wait would use about 60 of the company's 90 days before it had even been helped to apply.

The capacity limit falls out of the same inputs. At 0.4 proposals per partner and 20 reviews a month, the team supports 0.8 × 20 ÷ 0.4 = 40 enabled partners. Recruiting past 40 without new hours, shorter reviews or better proposals is recruiting a backlog.

| Month | Enabled partners | Proposals | Capacity | Utilization | Backlog at month end | Wait for a new proposal |
|---|---|---|---|---|---|---|
| 1 | 20 | 8 | 20 | 0.4 | 0 | None |
| 2 | 30 | 12 | 20 | 0.6 | 0 | None |
| 3 | 40 | 16 | 20 | 0.8 | 0 | None; pause recruiting here |
| 4 | 50 | 20 | 20 | 1.0 | 0 | None on average; any clumped week queues |
| 5 | 60 | 24 | 20 | 1.2 | 4 | About 1 week |
| 6 | 70 | 28 | 20 | 1.4 | 12 | About 3 weeks |
| 7 | 80 | 32 | 20 | 1.6 | 24 | About 5 weeks |
| 8 | 90 | 36 | 20 | 1.8 | 40 | About 9 weeks |

## What fills the queue: a low-fit cohort and repeat names (Illustrative)

Arrivals are not all alike. Break month 6 down by where its 28 proposals came from, and the queue turns out to be partly self-inflicted.

**The underperforming cohort.** Twenty partners recruited at one conference in months 4 and 5 sent 10 of the 28 proposals, and only 2 passed the fit screen, mostly because the companies had client-owned records or too few full-time staff. Each passing proposal from that cohort cost 7.5 review hours, three times the established partners' 2.5.

**The duplicate-risk case.** Three proposals named companies another partner had already claimed in your log, a pattern that grows as overlapping circles join: the accountant and the lender in one town know the same CEOs. A two-minute log lookup before the full review takes them out of the queue, about 4.5 hours saved. It cannot catch companies SourceX already knows through an earlier application, relationship, lead, account or email correspondence; those are duplicates for every partner, and only SourceX's records show them.

**Cheaper capacity first.** Asking the cohort to answer the five screening questions before review lets an obvious miss be returned in 15 minutes instead of 90, saving about 10 hours on its 8 misses. With the log lookup, month 6 needs about 27.5 hours against 30 available: under capacity, still above 0.8, so recruiting stays paused.

| Month 6 source (Illustrative) | Partners | Proposals | Passed fit screen | Review hours | Hours per passing proposal |
|---|---|---|---|---|---|
| Established partners | 50 | 18 | 11 | 27 | 2.5 |
| Conference cohort | 20 | 10 | 2 | 15 | 7.5 |
| Both, before fixes | 70 | 28 | 13 | 42 | 3.2 |
| Both, after log lookup and up-front screening answers | 70 | 25 reviewed | 13 | About 27.5 | About 2.1 |

## Capacity, support, concentration and overlap on one page

Run the model once a month alongside four checks, because each one moves either arrivals or hours. Support questions that are not about a specific proposal still eat review time, so measure them and subtract them from H before dividing. Concentration matters because one prolific partner's holiday can drop arrivals by a quarter and make the model look safer than it is; run it with and without your top partner. Overlap matters because duplicate proposals rise as networks touch.

Related companies are the overlap case nobody can settle in advance. Different partners may propose two subsidiaries of the same group; the group shares one reward cap, but whether an earlier introduction of one subsidiary affects credit for the other is not written into the terms. SourceX resolves it during duplicate review, so your log should record the group link and your team should promise neither partner anything.

When two partners collide over the same company, [seven evidence checks for one company and two partners](/guides/when-two-partners-refer-the-same-company) show which introduction counts first; more network playbooks are in the [guides hub](/guides).

| Check | Measure | Threshold (Illustrative) | Action |
|---|---|---|---|
| Support load | Partner questions not tied to a proposal, in hours a month | Over a fifth of H | Write one answer for the whole bench and subtract the rest from H |
| Concentration | Share of arrivals from the top partner | Over the greater of $50,000 plus 5% and 15% | Plan capacity on arrivals without that partner |
| Overlap | Proposals naming a company already in your log, per 100 | Over 5 | Make the log lookup mandatory before review |
| Cohort fit | Fit-screen pass rate per recruiting cohort, on at least 10 proposals | Under 40% | Up-front screening answers and a criteria session before any further recruiting from that source |

## Gates for the next recruiting tranche, and how to correct the model

Recruit again only when the evidence says the queue can absorb it. These gates are Illustrative; set your own before the next campaign, not after it.

The model will be wrong, and the correction path should be routine. Compare forecast and actual backlog each month; if they differ by more than a quarter two months running, re-measure p and h from dated records instead of arguing with the queue. Never fix utilization by telling reviewers to go faster: a shorter h bought by skipping the consent confirmation or the log lookup moves the cost onto CEOs and onto duplicates. When SourceX's criteria change, screen pass rates and review time change with them, because a screen is valid only on the date it was made, so tag every input with the criteria version it was measured under. To see which partners deserve the scarce review slots, pair this model with [the fit-trust-follow-through scorecard](/guides/the-partner-quality-scorecard-that-includes-company-feedback) and [behavioral partner segments](/guides/how-to-segment-partners-by-real-referral-behavior).

1. Utilization at or below 0.7 for two consecutive months.
2. No proposal waiting more than a week at month end.
3. The newest cohort's fit-screen pass rate measured on at least 10 proposals and at or above 40%.
4. No trust complaint from an introduced CEO in the quarter.
5. Inputs p and h re-measured within the last quarter, with the criteria version recorded.

## Where this model misleads

- **Averages hide clumps.** A webinar can produce a month's arrivals in a week. The 0.8 margin absorbs some of that, not all.
- **Little's formula needs a steady system.** Use it in stable months; in growing months, count the backlog directly.
- **SourceX's queue is invisible.** Its review times are unpublished, so the model ends where your team's work ends.
- **Pass rates are not acceptance.** A proposal that passes your screen can still be declined in review for relevance to current data needs or for rights.
- **No income belongs in it.** SourceX publishes no conversion rates or typical payouts, and a reward exists only after a completed deal and the payment conditions in the terms, so the model plans hours, not revenue.
- **Partner behavior responds to waiting.** Partners who wait weeks for a reply stop proposing, which lowers arrivals and makes a broken queue look healed. Read falling arrivals after a backlog as a warning, not relief.

## 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.

When your model shows room under the 0.8 line, have new partners [register for a partner link](/register) and sign before they propose anyone, screen each company against [who qualifies](/who-qualifies), and check the 90-day window in the [program terms](/terms) before the first introduction goes out.

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/who-qualifies
- https://pubsonline.informs.org/doi/10.1287/opre.9.3.383
