What would disprove your network's data opportunity? A disconfirming-evidence checklist that says whether a finding sinks one company, your network or the whole category

Short answer

Your network's data opportunity is disproven when the companies you can actually reach keep failing the same necessary condition: too small at peak, no multi-year records of their own, records that belong to clients, CEOs who will not talk, nobody in your network to ask, or companies SourceX already knows. Write each falsifier down before you look, fix how many hits it takes, and record whether a finding sinks one company, your network or the whole category. A single decline disproves nothing.

Visual briefFull-size diagram
A three-column board sorting disconfirming findings by scope: one company (a CEO says no, an archive deleted), your network (most never reached 50 at peak, records held for clients, nobody will ask, already known to SourceX) and the category (developers stop needing real records); the network column is highlighted.
Falsifiers from this page's checklist, placed by the scope each can sink. Placement follows the method on this page; no real network is shown.

The claim worth trying to break

"Our network is full of data" cannot be disproven, which is the best reason not to believe it. Rewrite it with a number and a date: "By the end of next quarter, at least five companies we can reach will pass the published fit facts, keep years of their own operating records and have a CEO whom someone of ours can ask." Now every company you look at either adds to that count or shows one of the ways the claim fails.

Three scopes matter, because the same finding means different things at each.

  • One company. A fact about a single business: it never reached 50 full-time employees at peak, its archive was switched off, or its CEO said no.
  • Your network. A pattern across the companies you can reach: most are young, most hold their clients' records, or nobody owns the CEO relationships.
  • The category. A claim about what AI developers need, which nothing your network observes can settle.

Most mistakes are scope mistakes: one enthusiastic CEO read as proof of a category, or a network-wide pattern of client-owned records read as bad luck.

From a market story to companies you could name

Market stories do not name companies. Epoch AI's 2024 paper, for example, projects that language models will fully use the stock of public human-generated text between 2026 and 2032 if current trends continue (our paraphrase). Even if right, it names no business in your network. Restate the opportunity at company level in four parts:

  1. The record. Cases, jobs or transactions tracked from start to recorded outcome, kept for years.
  2. The holder. A company doing that work on its own account, not a vendor hosting the software or a service firm working on clients' files.
  3. The size and place. 50+ full-time employees at peak, in a covered country.
  4. The path. A named person in your network who owns the CEO relationship and will ask.

A company that matches all four is a row on your list. The list, not the market story, is what the falsifiers test. If you have not built it yet, the network opportunity worksheet and the opportunity finder are the place to start.

The disconfirming-evidence checklist

Ten falsifiers. Each says how you would notice it without touching a record, what it can sink if it holds, and what it leaves standing.

#FalsifierHow you would see it, without touching recordsWhat it can sinkWhat it leaves standing
1Size: most reachable companies never reached 50+ full-time employees at peak, contractors excludedHeadcount history from the relationship owner or the CEOYour network, if it is the patternThe category, and larger companies you have not yet reached
2History: records start recently, or old systems were switched off without an archive"Since what year, in which system?", asked through the relationship ownerOne company, or your network if it repeatsA company whose archive survived elsewhere in its group
3Ownership: the valuable records are kept on clients' behalfThe CEO's account of whom the work is for and what client contracts allowOne company, or your network if that business model dominates itCompanies doing the same work on their own account
4Authorization: CEOs decline outrightThe CEO's answer, through the person who owns the relationshipYour network's opportunity this yearA CEO who said "not now"; a company after a change of CEO
5Reach: nobody in your network owns the CEO relationships, or the owners will not askYour relationship map, with each owner's answer recordedYour network as it standsThe companies themselves; a new connector can change it
6Already known: companies already have an application, account, lead or email history with SourceXHidden from you; it surfaces when an introduction comes back from duplicate reviewYour network as a source of credited introductionsThe companies' own chances of licensing
7Program fit: SourceX's current programs do not want the record types your network holdsSourceX's review outcome and any decline reason it givesThat record type, for nowOther record types, and later programs
8Language: companies outside the U.S., Canada and the EU whose offered records fall below 70% EnglishThe CEO's description of the language the records are kept inThose companiesCompanies whose records are kept in English
9Category evidence: credible, dated research shows developers no longer need records like thesePrimary sources: research papers and developer documentationThe category, and every network's version of the opportunityNothing, on its own, about one company's records
10Cost: relationship owners report that the asks spend more goodwill than they returnEach owner's feedback after an askYour network's appetite to continueThe opportunity inside someone else's network

Thresholds come first. For each falsifier, write the count that would make you accept it before you look at a single company. Illustrative defaults for a list of 20: accept falsifier 1 or 3 at 12 of 20 companies; falsifier 4 when 6 of the first 8 CEOs asked refuse outright; falsifier 6 when 3 of the first 5 introductions come back as duplicates. Falsifier 9 needs a source, not a count. Falsifiers 6 and 7 are the two you cannot observe yourself; SourceX's review supplies them, which is why they should be read from actual results rather than guessed. When a review declines a company, how to understand why a company referral was declined helps sort the reason into the right row. For the company-level stop signs inside each row, see the data licensing red flags screen.

The relationship and ownership facts behind each row

A falsifier can only be scored against a row that names who holds what. For each company, record five facts, each with a date:

FactWho supplies itWhy the checklist needs it
The relationship ownerYour own teamFalsifiers 4, 5 and 10 are about this person's ask
The CEOThe relationship ownerOnly the CEO can open licensing talks and sign for the company
The entity holding the records todayThe CEOAfter an acquisition the records may sit inside a buyer's group
What the records are, how far back, which systemThe CEO or the records owner the CEO namesFalsifiers 2 and 3
Peak full-time headcount and locationThe CEOFalsifiers 1 and 8

The published fit facts frame every row, and each one is a possible falsifier: 50+ full-time employees at peak, contractors not counted; an operating history the company can document; rights to license; and an authorized contact. 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. Questions stay at metadata level; nobody asks for files, samples or logins.

Groups add a wrinkle. For the reward cap, a whole corporate group is treated as a single company, however many of its entities license. What the published terms leave open is how earlier dealings elsewhere in a group play out: if a parent or sister business has already been in touch with SourceX, or a different partner has introduced a company from the same group, SourceX weighs the effect on credit during its duplicate review. Note such cases on the row rather than predicting the answer.

A three-week disproof sprint

Ask your relationship owners to list 20 companies they could raise the subject with, before anyone screens them. Over three weeks, collect the row facts at metadata level, record every falsifier a company hits, and read the results against the thresholds you wrote first.

Illustrative. A fictional sprint; no real companies or network.

Falsifier checked, in orderCompanies hitStill clearThreshold written firstReading
1. Size at peak91112 of 20Not accepted, but close
2. History2910 of 20Not accepted
3. Records held for clients4512 of 20Not accepted
5. No owner will ask1410 of 20Not accepted
4. CEO refuses outright1 of 4 asked36 of the first 8 askedToo few asked to read

No single falsifier crossed its threshold, yet the claim failed: three companies are clear of every falsifier, against the five it promised. A claim made of several necessary conditions can fail by accumulation even when each condition looks healthy on its own. The honest record says so, and then makes one of two choices: change the reach (recruit a connector into a part of the market the network does not touch) or accept that this network's opportunity, as written, is disproven for the quarter. Quietly lowering the target from five to three and calling it a success is the one move the method rules out.

When the checklist itself is wrong

  • Thresholds are judgment. Twelve of twenty is a default; the protection is writing it down first.
  • Two falsifiers are invisible to you. Duplicates and program fit come only from SourceX's review, so early in a sprint they are unknown, not passing.
  • Category evidence cuts both ways. Epoch AI's report on AI in 2030 argues scaling could continue for at least a few more years on existing public text and other kinds of data, while Shumailov and colleagues, writing in Nature in 2024, argue that access to original human-generated data matters when models learn from generated content (both our paraphrases). Neither mentions business records, so neither proves or disproves your network's version.
  • Rights. A CEO's account of ownership is not a rights review. Customer contracts and privacy limits are read by SourceX and the company's counsel.
  • Survivorship. Owners list the companies they are comfortable raising. Ask each which companies they left off, and why.
  • Programs change. A company that misses one SourceX program may fit another later, so a disproof dated October deserves a revisit date.

Methods note. The checklist was built on 2026-10-26 from the program terms (version 2026-10-22.2, since replaced by version 2026-11-04), SourceX's published screening criteria and the research summaries cited above. It measures whether a network's reachable companies meet necessary conditions; it does not measure demand for any company's records, record quality or any payout, which only SourceX's review can assess. The sprint figures are fictional; no dated mandate feedback or founder interviews have been published for a real network, so those figures are not yet available.

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 a company survives every falsifier and its CEO wants the conversation, register for a partner link, check the row against who qualifies and read the program terms before the relationship owner makes the ask. For sharper questions to put to the survivors, use how to turn a contrarian data thesis into testable questions; for the wording of any note about them, see the difference between a category bet and a company endorsement, or browse the other partner guides.

Trying to disprove a network's opportunity is referral work; whether SourceX itself deserves investment is a different test, and investor materials for it can be requested through 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

Does one CEO declining disprove a network's data opportunity?

No. A single decline is a fact about one company, and often about one moment: a CEO who says not now may say yes after a migration or a change of leadership. Only a pattern you set a threshold for in advance, such as most CEOs on your list refusing outright, says something about the network.

Can the opportunity fail for our network while the category still holds?

Yes, and that is the most common result. A network can be too young, too small at peak or built around firms that hold their clients' records, while companies elsewhere keep exactly the records AI developers want. Record it as a finding about your reach, not about the market.

How do we find out that SourceX already knows a company?

You usually cannot see it in advance. SourceX checks its own records before inviting an introduced company, and a company with any earlier lead, account or email exchange is reviewed as a duplicate. You learn it when an introduction comes back from duplicate review.

Sources

  1. refer.sourcex.si/terms
  2. refer.sourcex.si/who-qualifies
  3. epoch.ai/publications/will-we-run-out-of-data-limits-…
  4. epoch.ai/files/AI_2030.pdf
  5. nature.com/articles/s41586-024-07566-y
  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