The client problem, and how small the partner's part is
Start with what the client gets, because a channel that does not help clients will not survive its first awkward conversation. Some client companies hold years of their own operating records: support histories, project files, quality records, finance workflows. SourceX takes approved business data under license straight from the companies that created it, for AI training and evaluation, and handles everything around that: the agreement, the rights review, data preparation, delivery and payout. For the right client, that is a new and optional use of records it already keeps.
The partner's part is narrow. Your firm recognizes a client that might fit, asks the CEO whether they want a conversation and, if they do, submits the introduction. You never handle the client's records, never describe their contents beyond metadata (what exists, roughly how much, how many years, which system, who owns it) and never negotiate for them. The CEO authorizes the licensing discussion and signs for the company.
SourceX names the professionals best placed to refer: M&A advisors and business brokers, private equity and operating partners, fractional CFOs and accountants, IT managed-service providers, wind-down and restructuring advisors, and operators and executives. The same page tells licensed professionals to check their own rules on referral fees. Before scoring anything, read referral fee rules by profession, and if you are weighing a bigger move, whether advisors should refer or build a practice.
Why the first introductions are the wrong evidence
A firm's first introductions are not a sample. They are a selection. The partner who signed up chose the clients they know best, whose CEOs return calls and whose fit was obvious. Of course those went well. The expansion question is about everyone else.
So the first thing to measure is the size of what is left. Illustrative: an advisory firm with 140 client companies screens its book at metadata level.
| Screen step (Illustrative) | Clients remaining |
|---|---|
| Whole client book | 140 |
| Reached 50+ full-time employees at peak, contractors excluded | 38 |
| Also keep several years of their own records, not records held for their clients | 24 |
| Also have a CEO someone at the firm can ask | 15 |
| Not among the first 4 already introduced | 11 |
The firm's four early introductions all went smoothly. That tells it little. The 11 are the expansion. If the last row had read 2, no early success would justify a firm-wide program, and the honest answer would be to keep the channel small.
Method note. Every test in the screen comes from two published pages, the who-qualifies page and Version 2026-10-22.2, since replaced by version 2026-11-04 of the program terms, both checked 2026-10-26. It counts only what the firm can confirm from its own knowledge or from the client, and treats unknowns as unknown rather than as passes. It does not estimate demand for any record type or any payout; SourceX's review decides program fit, and SourceX has not published conversion rates or typical payouts.
The remaining-pool expansion scorecard
Four scored lines, four pass-or-fail gates and one line that is recorded but never scored. Score after at least three resolved asks.
| Line | What you count | Where the count comes from | 0 | 1 | 2 |
|---|---|---|---|---|---|
| Remaining pool | Unasked clients passing every fit fact you can confirm | Your metadata screen | One or none | Two to nine | Ten or more |
| Client usefulness | Clients asked who said the conversation was worth their time, including those who declined | One question asked 30 days after each conversation | Fewer than half | Half or more | Half or more, plus at least one unprompted request |
| Hours per agreed introduction | Staff hours divided by CEO-agreed introductions | Time logs | Above the limit you set before starting | At or below it | At or below it and falling |
| New to SourceX | Introduced companies that were not already known to SourceX | SourceX's duplicate findings | Most were duplicates | Some duplicates | None or one |
| Gate | Passes when |
|---|---|
| Consent | Every submitted introduction had the CEO's agreement first |
| Professional rules | Someone has checked your profession's rules on referral fees in writing and the client disclosure is settled |
| Metadata only | No one at the firm requested or handled a client's records, samples or logins |
| Second screener | A colleague screens five clients independently and reaches the same fit result |
Recorded, never scored: cash. Rewards actually paid, with dates. Estimates, proposed amounts and calculator figures stay out.
Decision rules, written before scoring:
- Decline when a gate fails and cannot be fixed, or when both the remaining pool and client usefulness score 0.
- Pilot when fewer than three asks have resolved; the scorecard has nothing to read yet.
- Keep it as it is when all gates pass but the remaining pool scores 1, or the total is 3 to 5.
- Expand when all gates pass, the remaining pool scores 2, client usefulness scores at least 1 and the total is 6 or more.
The companion guide on keeping a referral program small covers how to run the channel at its current size; this scorecard decides whether it should change size at all.
Two participation models, side by side
Expanding usually means moving from the first model to the second. Compare them on what the client experiences, not on what the firm might earn.
| Question | Model A: relationship-led | Model B: firm-wide screen |
|---|---|---|
| Who spots a fit | One or two senior partners, from their own clients | Staff run a short metadata screen during regular client reviews |
| Who asks the CEO | The partner who owns the relationship | Still the relationship owner, never the screener |
| Client experience | Raised when it plainly fits | Every client in scope is screened; only fits hear about it |
| Firm effort | A few hours per introduction | Training, a screen template, a review step and time logs |
| Main risk | The pool runs dry after the obvious fits | Staff drift from metadata into records, or the topic reaches clients it does not suit |
| What must be true first | Gates pass | Gates pass, the remaining pool scores 2 and a second screener agrees |
Model B is not a bigger version of Model A. It adds a screening step, and every added person is another chance for discovery to slide from "how many years of tickets exist" to "send me a sample." The metadata gate matters most here.
Hours, fixed terms, client trust and unknown economics
Effort is the line firms underestimate. Count the screen, the CEO conversation, the follow-up and any internal review. If you set an hours limit before starting, the scorecard can hold you to it.
Terms are fixed and public. The firm earns the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap, on a referred company's first qualifying transaction; further qualifying transactions with that company first agreed within 12 months of the first Closing pay 15% of their Actual Supplier Payout, the USD 50,000 floor applies once per referred company and does not return on a repeat, renewal or amendment, and a transaction qualifies only when its Actual Supplier Payout is at least USD 100,000 unless SourceX agrees in writing. The base is never SourceX's fees or revenue; SourceX pays the reward itself, separately, and nothing comes out of the client's payout. Illustrative: if a referred company's first qualifying Actual Supplier Payout were $250,000, the reward would be $62,500; at $400,000 it would be $70,000, with no cap above that. Five conditions must all hold before it is payable: the Acquisition Partner has accepted, SourceX has received cleared payment, the company's compensation has been paid, SourceX finance has approved, and your tax and payment details are on file. Payment is due within 60 calendar days of whichever happens later: the Acquisition Partner gives final written acceptance of the complete, validated dataset, or the handoff of that dataset is completed; if a condition is still open on that date, the reward follows promptly once it is met and no fresh 60-day period starts. No credit goes to a duplicate company, meaning one that already had an application, relationship, lead, account or email correspondence with SourceX.
The agreement itself is a modest commitment: it runs for one year and may be ended on 30 days' notice, but qualifying contacts stay protected for six months after it ends and confidentiality lasts three years, longer for trade secrets. Expanding can be reversed; the obligations do not vanish with it.
Trust is the line clients feel. A client who hears about a data program from their accountant during a tax review may wonder whose interest is being served. That is why the usefulness line counts decliners and why the professional-rules gate comes before any scoring.
Uncertain economics cut both ways. Large partner programs struggle to see their own results too: in a PartnerStack and Wynter survey of 100 senior leaders at B2B SaaS companies with $50M+ revenue, 14% named lack of visibility into partner impact as their biggest obstacle, behind team alignment (37%) and having no clear partner program (20%). That population is not advisory firms, but the lesson transfers: a channel nobody measures cannot be judged, so the scorecard is the measurement.
Pilot, keep, expand or decline: three firms (Illustrative)
Illustrative. Generic firms and invented scores; no outcomes implied.
| Firm (Illustrative) | Remaining pool | Usefulness | Hours | New to SourceX | Gates | Decision |
|---|---|---|---|---|---|---|
| 12-person accounting practice | 1 (three clients left) | 2 | 1 | 2 | All pass | Keep as it is |
| 60-person advisory firm | 2 (eleven clients left) | 1 | 1 | 2 | All pass | Expand to Model B in one practice group |
| Managed IT provider | 0 (most clients below the size line) | 1 | 0 | 1 | Metadata gate failed | Decline |
The accounting practice scores 6 but has only three clients left to ask, so the pool line holds it at its current size. The advisory firm passes every rule and widens carefully, one practice group first. The IT provider's technician offered a client's export "to show what they have," which fails the metadata gate; combined with a thin pool, the right answer is to decline and write down why. For accounting firms specifically, see the CPA firm advisory partner page.
When the remaining-pool score points the wrong way
- Small numbers. Three or four resolved asks make every line noisy. Treat scores as direction and rescore each quarter.
- The pool estimate depends on what you know. Clients you know least about will look like fails or unknowns. Ask before you rule them out.
- Duplicates arrive late. The novelty line fills in only when SourceX has checked its records, so early scores may look better than they are.
- Program fit is not yours to score. A client that passes every fit fact can still fall outside SourceX's current data needs; the company fit checker gives a non-binding first pass, and review decides.
- The survey is about software companies. The PartnerStack figures describe large SaaS firms, not advisory practices, and come from the publisher's own research.
For the client's side of the same question, the main site explains what company data may be worth to AI developers, and whether licensing company data is worth it answers it from the owner's chair.
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 scorecard says pilot, keep or expand, register for a partner link, screen the remaining clients against who qualifies and read the program terms before any relationship owner raises it with a CEO. Related channel decisions are collected in the guides library.