What would prove a venture network is becoming more useful? A falsifiable network-value scorecard with six measures, fixed denominators and the result that would count against each one, written down before the first quarter is counted

Short answer

Only a result that could have come out the other way. A venture network is becoming more useful when, quarter after quarter, a larger share of its asks give the companies it reaches a quick, clear answer about a real fit, with fewer wasted asks and no rise in complaints, all measured with definitions fixed in advance and with unanswered, declined and failed asks kept in every denominator. This page publishes that test: six measures, a comparison rule and the results that would count against the claim. No network's results are published here.

Visual briefFull-size diagram
A schematic chart with no data. A lock icon marks the plan date on the left, followed by four empty quarter columns from baseline to the read date. Above a dashed baseline line, a green band is labeled would support the claim; below it, an orange band is labeled would count against it. A note reads no results published.
Schematic only: the bands show how results will be read, not any result. No network has been measured with this scorecard yet.

The question matters to a platform research director because "our network got more useful" is the sentence every fund wants to say and almost none can check. Activity counts rise whenever a team works harder. Usefulness, as defined below, can fall even while activity rises, which is exactly what makes it worth measuring.

Useful to whom, and what counts as one ask

The scorecard measures usefulness from the company's side, not the fund's. A network is useful when the CEOs it contacts get something from the contact: a relevant question, asked once, answered quickly, leading somewhere when the answer is yes.

TermDefinition used on this page
AskOne relationship owner asking one company's CEO whether they want a conversation about SourceX, logged with its date
Relationship ownerThe person in the network who holds the tie to that CEO: a partner, platform lead, LP, portfolio founder or advisor
Fit-passingBy the CEO's own answers, the company meets every published fit fact the network can check before review: 50 or more full-time employees at peak with contractors left out, years of its own records, and location
Clear answerA yes or a no from the CEO; silence is not an answer
Follow-throughThe company submits its own application within the 90-day window that section 2 of SourceX's terms sets for credit
PeriodA calendar quarter; the baseline is the first full quarter after the plan is locked
Harm signalA complaint, an opt-out or a request not to be asked again

On location, the published sentence is the test: 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. SourceX also screens each company against the criteria of a specific program, and those can move, so fit is scored against the published facts as they stood on the date of each ask.

The falsifiable network-value scorecard

Six rows, each with its numerator, its denominator, the direction that would support the claim and the result that would count against it. Row 6 is a veto.

RowMeasure (numerator over denominator)Supports "more useful"Counts against itKnown confound
1. Fit precisionFit-passing asks over all asks, with unknown fit counted as not passingRisesFlat or lowerProgram criteria change between quarters
2. Clear answersAsks with a CEO yes or no within 30 days over all asksRisesFallsFewer asks sent to busy CEOs
3. Follow-throughCompanies that applied within 90 days over CEO yesesRisesFalls while yeses riseA few owners with unusually strong ties
4. Reach beyond the portfolioFit-passing asks to companies outside the funded portfolio over all fit-passing asksRisesUnchanged or fallsA new fund vintage changes the portfolio
5. EffortPlatform hours over CEO yesesFallsRises two quarters runningHours logged less carefully over time
6. Harm (veto)Asks drawing a harm signal over all asksStays flat or fallsAny rise, whatever rows 1 to 5 showComplaints that never reach the network

Row 3 is the one most likely to embarrass a good-looking quarter. If CEO yeses climb but follow-through drops, the network is collecting polite agreement, not useful conversations. Row 4 tests a thesis rather than a fact: SourceX's screening guidance says funding plays no part in its checks and that the useful venture channel often runs wider than the funded portfolio. If the share stays flat, the network is not finding the companies its cap table cannot reach.

Completed licenses and rewards are deliberately absent. They depend on SourceX's review, the company's own choices and an Acquisition Partner's acceptance, and a single network sees too few of them in a quarter to read anything.

The comparison rule, locked before the first quarter

Write the rule down, date it and file it somewhere the team cannot quietly edit. The Center for Open Science describes preregistration as "specifying your research plan in advance of your study and submitting it to a registry", and says it helps tell planned work from unplanned work. A fund need not use a public registry, but it needs the same discipline.

The rule this page proposes:

  1. The claim holds only at the read date, comparing the fourth quarter with the baseline, never on one good quarter in between.
  2. At least four of rows 1 to 5 move in the supporting direction by the margin written into the plan (for example five percentage points for a share, a tenth for hours).
  3. Row 6 does not rise. A rise ends the claim regardless of the other rows.
  4. Each quarter holds at least 30 asks. With fewer, the report says "not enough asks to tell" rather than guessing.
  5. Portfolio and non-portfolio asks are reported separately. If the improvement appears only in the pooled figure, it came from a change in mix, and the claim fails.

No row is judged on a significance test alone. The American Statistical Association's 2016 statement puts it directly: "Scientific conclusions and business or policy decisions should not be based only on whether a p-value passes a specific threshold." With samples this small, the margins, the counts and the counterexamples carry the weight.

Frame. Every ask made by relationship owners who agreed to take part, from the day the plan is locked. Asks made before that day stay out. This is a census of the network's own activity, not a sample, so nobody chooses which asks to count.

Consent and minimization. Relationship owners agree to log their asks. CEOs are not surveyed; the log records only what the network already knows from the ask itself. No company records, files or system details are collected, and partners never handle client or company data. Company names stay in the platform team's log under codes, and published tables suppress any cell under five.

The ask log. Eight fields per ask: company code, relationship owner code, ask date, portfolio or not, fit facts as the CEO stated them, the CEO's answer and its date, application status (from the company or the partner dashboard, which shows each referral's stage), and any harm signal. Hours come from one quarterly total per person, not a new diary.

Missing records. An ask with an unknown outcome stays in every denominator and is reported as unknown. A relationship owner who logs nothing for a quarter is listed, because silence in a log is more often missed logging than missed asking.

Descriptive results, not causal ones

Even a clean pass would show only that the network's asks improved, not that the network caused the improvement. Four rival explanations must be ruled out in the report, each with its own check:

  • The program changed. If SourceX's published criteria moved during the year, row 1 is rescored against the facts as they stood at each ask date.
  • The people changed. New relationship owners with stronger ties can lift every row; report results with and without owners who joined after the baseline.
  • The baseline was a bad quarter. A team that started measuring after a poor stretch will see improvement whatever it does; the plan names the baseline before anyone sees its numbers.
  • The calendar. Quarters differ in how reachable CEOs are; compare the read quarter with the same quarter a year earlier once two years exist.

For the separate questions of whether individual partners improve with practice, and how dependent a network is on a few people, see the learning curve study and the network concentration report. A network too small for quarterly rows can start with the ten-introduction experiment.

What the published report will contain

The report publishes the locked plan and its date, then this table with every cell filled, including the rows that went the wrong way:

RowBaselineQuarter 2Quarter 3Quarter 4Direction met?
1. Fit precisionNot yet observedNot yet observedNot yet observedNot yet observedNot yet observed
2. Clear answersNot yet observedNot yet observedNot yet observedNot yet observedNot yet observed
3. Follow-throughNot yet observedNot yet observedNot yet observedNot yet observedNot yet observed
4. Reach beyond the portfolioNot yet observedNot yet observedNot yet observedNot yet observedNot yet observed
5. EffortNot yet observedNot yet observedNot yet observedNot yet observedNot yet observed
6. Harm (veto)Not yet observedNot yet observedNot yet observedNot yet observedNot yet observed

Alongside it: counts and denominators for every cell, the portfolio and non-portfolio split, one anonymized counterexample per row (an ask that moved against the trend and why), every deviation from the plan with its date, and the limitations below. Results are not published on this page, and none exist for this design yet. Any statement about SourceX's own outcomes belongs to SourceX's verified records, as partner results research sets out.

Results that would sink the claim, and results that would sink the scorecard

The claim fails if the rule above is not met. The scorecard itself needs rethinking if any of these appear:

  • Rows 2 and 3 move in opposite directions for two years. Quick answers and real follow-through may not belong in one score.
  • Row 6 stays at zero while relationship owners report awkward conversations. The harm signal is too blunt and needs a better source.
  • Fit-passing companies decline at the same rate as misfits. Then fit precision does not track what CEOs find useful, and the definition of usefulness is wrong.

Risks and limits of the method

  • Small numbers. Most networks will make dozens of asks a quarter, not hundreds; margins and counterexamples matter more than any rate.
  • Self-reporting. Owners log their own asks; the missing-records list exposes gaps but cannot close them.
  • Dependence on SourceX. Row 3 relies on application status the network learns from the company or the dashboard, and program criteria can change during the year.
  • Narrow definition. The scorecard measures usefulness for data introductions only, not a network's value for hiring, sales or fundraising.
  • No benchmark. No published figures exist to compare a network against, and this page supplies none.

Methods note. The definitions and fit facts follow SourceX's terms (version 2026-10-22.2, since replaced by version 2026-11-04), its screening page and its process page, all read on 2026-10-28; the comparison rule follows the preregistration practice the Center for Open Science describes and the ASA's statement on p-values. The six rows, margins and minimum counts are this page's design and have not been tested on a real network.

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.

A network that wants its own asks in the log should register as a partner and sign the Channel Partner Agreement before submitting anything, check each company against who qualifies, and keep the program terms version beside the locked plan so the definitions match. More method pages sit in the guides library.

Measuring a network's usefulness is a referral question; judging SourceX as a possible investment is a different exercise, and anyone doing that can request materials from 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.

  1. Step 1Share your linkSend your personal link to a company you know.
  2. Step 2Company appliesThe company applies itself at /apply.
  3. Step 3Company signs and is paidThe company signs, delivers its data and is paid by SourceX under its agreement.
  4. Step 4You get your rewardthe greater of $50,000 plus 5% and 15% of the actual supplier payout, with no cap. A reward becomes payable after SourceX has paid the referred company under its signed agreement and the payout conditions in the program terms are met, including finance approval and your tax and payment details.

Common questions

Why not just count introductions or registered partners each quarter?

Because both can rise while the network wastes more founders' time. A count with no denominator cannot fall for the right reasons, so it cannot show improvement either. The scorecard keeps every ask, including the ones that went nowhere, and asks what share of them helped the company.

Do completed licenses or rewards appear on the scorecard?

No. They arrive months after the ask, depend on SourceX's review and the company's own decisions, and are too rare in one network's quarter to read. The scorecard stops at the company's own application, the last step a network can fairly claim to influence.

Can a fund publish its own results using this method?

Yes, if it locks the plan first, keeps every row including the ones that move the wrong way, and reports only aggregated counts above the minimum cell size. Any outcome it cites about SourceX itself should come from SourceX's own records, as the partner results research page explains.

Sources

  1. cos.io/initiatives/prereg
  2. amstat.org/asa/files/pdfs/P-ValueStatement.pdf
  3. refer.sourcex.si/terms
  4. refer.sourcex.si/who-qualifies
  5. refer.sourcex.si/how-it-works
  6. sourcex.si/investors

Free resources

Know a company with valuable proprietary data?

Become a referral partner, get your link and introduce the company’s CEO (in an owner-run company, usually the owner). Referral partners can join from most countries. We can't work with partners in countries or regions under U.S. sanctions or on U.S. restricted-party lists. The companies you introduce should be based in the U.S., Canada and the EU, or in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia with at least 70% of their data in English.

Refer a company →

I own a business

Explore licensing your company's data directly to SourceX for AI training and evaluation. Start a short assessment; no uploads needed.

Start an assessment