How to publish network research without exposing companies: a twelve-gate anonymization and consent review that tests every table, quote, chart and footnote for a route back to a company, and holds the release until two intruders fail to name one

Short answer

Treat every number, quote and chart as a possible route back to a company, and close each route before release. Use only the network's own records and what each company agreed to share; strip names and anything searchable; coarsen combinations of sector, size, region and dates; hide small cells and a second cell beside each, so totals cannot be subtracted; keep declined companies in coarser rows instead of dropping them; then have two people try to name the companies. Publish only when both fail. This page reports no research results of its own.

Visual briefFull-size diagram
An Illustrative draft row on the left, with four lines struck through in orange (a named dental group in a named city, staff count and founding year, the introducing office and month, and a declined status with SourceX's reason), passes through four stacked gates, consent tier, names and combinations, cells and totals, and a two-intruder trial, and comes out on the right as a green published row reading healthcare, Midwest, first quarter, 12 asked, 9 ended with no application.
Illustrative row. The company, the counts and the gates' effects are invented to show the method.

Whose information a network report may hold

A platform research team usually holds three kinds of information, and only two of them are its own to publish.

InformationExampleStatus
The network's own activityAsks made, dates, stages the partner recorded, partners' own stepsThe network's own, subject to the gates below
What a CEO agreed to shareA quote, a case outline, the company's namePublishable in the form and scope the CEO approved
What SourceX told a partnerStage updates on an introduction, reasons for an outcomeConfidential unless SourceX authorizes publication in writing

SourceX's terms (section 7) state that "Partners never receive access to other partners' records, applicant documents or confidential buyer information," and section 8 leaves confidentiality to the signed Channel Partner Agreement. The unsigned sample of that agreement says: "Neither Party may publicly identify the other as a partner or disclose non-public deal details without written authorization, except as legally required." It also describes the updates partners receive as generic stages or reference codes. Treat those updates as non-public deal details and ask SourceX in writing before anything drawn from them appears in print. SourceX's own rule for citing results, permissioned and verified records only, is set out in partner results research.

Reporting inside a fund, where names never leave the team, follows a different design: a monthly investor partner report that protects confidentiality. A public release is the harder case, because its readers include people who know the companies.

Rewards are a special case. A reward is the greater of $50,000 plus 5% and 15% of what SourceX actually paid the company, so a published reward below the cap publishes that company's payout. No network report should carry one.

TierWhat may appearConsent neededRecord
NamedThe company's name with agreed factsThe CEO's written consent to the exact final text, plus SourceX's written authorization for anything about its dealDated message with the approved text attached
Described, unnamedA case with coarsened factsThe CEO's consent to anonymized useDated message naming the case
Counted onlyThe company inside totalsNone beyond the network's own record keeping, provided no cell exposes itThe frame list

People need their own consent: every partner or CEO quoted approves the final wording. Consent can be withdrawn until release, and the methods note names a contact for removal requests afterward.

Run all twelve gates on the complete release: tables, text, charts, appendices, file names, alt text and the spreadsheet behind them.

GateQuestionPass whenFix when it fails
1. SourceDoes every figure come from the network's own records or consented material?Each table names its sourceRemove it, or obtain written authorization
2. Consent tierIs every company in a tier with a dated record?The tier log is completeMove the company to counted only
3. Direct identifiersDo names, domains, products, systems, people, places or logos remain?None, file names and alt text includedStrip or genericize
4. CombinationsDo sector, size band, region, founding period and dates together describe fewer than 10 companies in the frame?Every published combination covers 10 or moreWiden bands, merge regions, use quarters instead of dates
5. Small cellsDoes any cell rest on fewer than 10 companies?None, or each is withheldWithhold it and fold it into a higher row
6. SubtractionCan a withheld cell be recovered from a total and the visible cells, or from last edition's table?Two or more cells withheld wherever one is, and editions comparedWithhold a second cell, or publish the total alone
7. DominanceDoes one company drive a sum, an average or a maximum?No single company makes up most of any figureReport counts, not sums; drop maxima
8. Outcome exposureCould a reader attach a declined, duplicate or withdrawn status to a guessable company?Outcomes appear only at levels that pass gates 4 and 5Merge outcomes, for example ended without an application
9. Introducer routeDo partner, office or fund names plus a sector point to one company?Introducers appear by role onlyRemove office and fund labels
10. Quotes and storiesDoes a quote carry a distinctive phrase, event or number?Each quote is consented in its final formParaphrase or cut
11. Graphics and filesDo maps, labeled points, metadata, hidden columns or version history leak?File checked and metadata strippedRedraw or re-export
12. Intruder trialCan either intruder name a company within an hour?Both failReturn to gate 4

Gates 5 and 6 borrow a habit from statistical agencies that publish figures about companies. Census Bureau tables mark a figure D: "Withheld to avoid disclosing data of individual companies; data are included in higher level totals." The Bureau of Labor Statistics likewise withholds detailed employment and wage figures where needed to protect the identity of cooperating employers, while its higher-level totals still include them (our paraphrase). Hide the cell, keep the company in the total. The threshold of 10 is this page's choice, the same one used in the research blueprint for company data introductions; no regulator set it for network reports.

The two-intruder trial

The UK Information Commissioner's Office describes a motivated intruder test for anonymized information about people: assume someone reasonably competent, with resources such as the internet and public records, who may ask people who know the subject (our paraphrase). This page adapts the idea to companies, which the ICO's guidance does not address.

  1. Intruder A knows the network: a colleague outside the research team who has never seen the frame list.
  2. Intruder B knows only public sources: company websites, registries, press and professional profiles.
  3. Each gets the full draft and one hour, and writes down every company they believe is described, with a confidence level.
  4. The research lead checks the guesses against the key. A correct name sends the release back to gate 4. So does a confident wrong name, because a reader could attach a declined status to a company that was never asked.
  5. The methods note publishes the trial log: companies described, guesses made, correct names, confident wrong names and the fixes made.

Declined and unfinished cases stay in, coarsened

When a gate fails on a declined company, the temptation is to drop the row. Dropping it removes a failure from the denominator, and the report then flatters the network. Fold it into a broader row instead, the way agency tables keep withheld figures inside their totals. Merging has a cost the report must state: a table that combined three sectors to pass gate 4 cannot support a claim that one sector converts better, and a merged outcome category cannot say why companies stopped. Descriptive counts survive anonymization better than causal comparisons; say which the release offers. For reports split by office or region, reporting international referral results fairly shows the row structure that the gates then coarsen.

One row through the gates (Illustrative)

StageRow as it stands
DraftDental group in Columbus, 140 staff, founded 1998, introduced by our Ohio office in March, declined after SourceX's rights review
After gates 1 and 3Dental group, Midwest, 140 staff, founded 1998, introduced in March, ended without an application
After gates 4, 5, 8 and 9Healthcare services, Midwest, first quarter: 12 asked, 3 applications, 9 ended without an application
After gate 12Unchanged: both intruders failed to name a company

The reason for the outcome left at gate 1, because it was SourceX's information. The city, staff count, founding year and office left at gates 3 and 9. The single company became one of 12 at gates 4 and 5, and its status merged with others at gate 8. A partner who wants to turn a single case into a teaching example has a separate protocol in what a declined referral can teach the next partner.

Testing the checklist itself

Whether the twelve gates work is an empirical question, and no answer exists yet. The test: collect draft releases from participating networks over a year; run the intruder trial on each draft before the checklist and again after it; count correct names and confident wrong names per 100 companies described; compare each draft with itself. Nothing is reported until 10 drafts have passed through both trials. If drafts that cleared the checklist still yield correct names, the thresholds rise from 10 to 20 or the combination fields shrink. If drafts yield no names even before the checklist, the gates cost more than they protect for that kind of table, and the report says so.

Limits a careful release still carries

  • Thresholds are judgment. Ten companies per cell is a working rule, not a standard, and a determined insider can beat it.
  • Small networks. A network that asks 30 companies a year may never publish a sector breakdown, and should not.
  • Other publications. A press story or another network's report can supply the missing piece. Gate 6 compares your own editions, not the world's.
  • Scope. The checklist governs what your network publishes. It says nothing about how SourceX handles any company's data, and no page of yours should imply otherwise.

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.

Research that passes the gates can sharpen where your network looks next; when it points to a company whose CEO wants the conversation, register as a partner, set that company beside the published fit facts on who qualifies, and keep the program terms at hand while its CEO decides whether to talk. More study designs are in the guides library.

Publishing research on your own network is separate from assessing SourceX as an investment; readers weighing the second can request investor materials 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

Can a network publish the referral rewards its partners received?

Not without exposing companies. A reward is a fixed share of what SourceX paid the company, so a reward below the cap reveals that payout, and SourceX's sample partner agreement bars disclosing non-public deal details without written authorization.

Is removing company names enough for a short case study?

Rarely. Sector, size, region, founding year and a dated event together often point to one company in a small network. Coarsen those fields, get the CEO's consent to the final text, and run the intruder trial on the case before release.

What if a company asks to be removed after publication?

Have a named contact and a takedown route in the methods note, remove the company's case or quote promptly, and re-run the subtraction gate, because pulling one row can expose another through the totals.

Sources

  1. refer.sourcex.si/terms
  2. sourcex.si/documents/partner/SourceX_Channel_Partner_Ag…
  3. census.gov/data/developers/data-sets/Annual-Survey-of-M…
  4. bls.gov/cew/overview.htm
  5. ico.org.uk/for-organisations/uk-gdpr-guidance-and-resou…
  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