How to state data language without guessing from headquarters: a language evidence note that gives each record stream its language, share, basis and source

Short answer

Headquarters tells you which language rule may apply, not what language the records are in. State language per record stream: the language, written as a standard tag such as en-IN or es-419, the share the company reports, how that share was measured, who measured it and on what date. Mark what is estimated or unknown, keep notes on translation and mixed-language content, and let SourceX measure the dataset actually offered. No records leave the company.

Visual briefFull-size diagram
A six-rung evidence ladder for a language claim, from a headquarters address at the bottom through website, staff location, interface settings and the company's own count to SourceX's measurement at the top; only the top two rungs can carry a share.
The evidence ladder behind the language note. A share needs the company's own count or better.

The published rule, word for word: 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.

Read it closely. The company's location picks which half of that sentence applies. The 90% in the second half is a property of the data, and no address can supply it.

What an address can and cannot tell you

A portfolio operations analyst filling in a referral summary is tempted to write "Language: English" because the company is in Singapore, Dubai or Bengaluru and does business in English. That line reads as a fact and rests on nothing.

  • Official languages are plural. Singapore's Constitution names four official languages: Malay, Mandarin, Tamil and English. A Singapore address does not say which of them a given team typed its notes in.
  • Regions are not languages. "Latin America" in the rule is a region of many countries, not a language. The standard language tags even give Spanish as used across Latin America and the Caribbean its own region code, es-419.
  • No share is not no language. For a company in the U.S., Canada or the EU, the published rule sets no English share. SourceX's who-qualifies page still says its current data needs differ in language, so a Montreal company's French service notes or a Munich company's German engineering reports are worth stating plainly.
  • Head office and records can sit apart. A company headquartered in one country may create its records in another. Write both countries and let SourceX decide in review; the published rule does not say how such a group is treated, and sorting entities is covered in checking language coverage before suggesting global fit.

Context to gather before any language line

Five facts per record stream, collected through whoever holds the CEO relationship, in the company's words: the system and record type; the entity and country that created it; the years it covers; which fields are free text and which are codes, numbers or dates; and the administrator who knows it. Add the CEO as the sponsor and whether the CEO wants a conversation.

The free-text question is the one most summaries skip. Language lives in narrative fields, documents and audio. A ledger of amounts, dates and product codes carries very little of it, while exception notes, email threads and call recordings carry nearly all of it. A language line that does not say which fields it describes cannot be checked.

The language evidence note template

One note per record stream. The company fills it in; the analyst checks that every field has an answer, including "unknown".

FieldWhat to writeThe weak version to avoid
1. StreamSystem and record type, with years: "exception notes in the transport management system, 2016-2026""Our data"
2. Made byThe entity and country whose people created itThe head office city
3. Content typeFree text, structured codes, documents or audio, and the mixNothing, as if all fields were alike
4. Languages, taggedStandard tags, such as en-IN, ar-AE or es-419"Mostly English"
5. ShareA range per language for this streamOne overall figure for the whole company
6. BasisThe evidence rung the share rests on (see the ladder below)Blank
7. Mixed or translatedTemplates in one language with notes in another; anything translated, when and howSilence
8. Supplied byRole, date, and whether the CEO approved the wording"The company says"
9. StatusReported, estimated or unknownA confident number with no status

The tags in field 4 come from the IETF's BCP 47, published as RFC 5646. It describes region subtags as marking variations associated with a specific country, territory or region, which is exactly the distinction a summary needs: en-IN says English as used in India, de-AT German as used in Austria. Writing the tag forces the question "which language, as used where?" and does not claim anything the company did not say.

The evidence ladder for field 6. Each rung supports a different kind of statement.

RungBasisWhat it can support in the note
0Headquarters addressNothing about language; write "unknown"
1Website and marketing materialWhich audience the company writes for; "unknown, likely English" at most
2Where staff and customers areA hypothesis to put to the company
3Interface language settings of the systemWhat the screens display, not what people typed into them
4The company's own count on a sample spread across years and teamsA reported share, labeled as the company's estimate
5SourceX's measurement of the dataset actually offeredThe share that counts

Only rungs 4 and 5 put a number in field 5. Rungs 0 to 3 justify a question, never a share.

Method note. The template and ladder were built for this page from the published geography sentence and the language variation noted on SourceX's who-qualifies page, with tag notation from RFC 5646, all checked 2026-10-27. The note records what a company reports and how it knows it. It does not measure record quality, demand for the records or rights to license them, and it does not replace SourceX's own measurement.

One freight forwarder, two notes (Illustrative)

Illustrative. A generic company with invented figures; no outcome is implied.

A 240-person freight forwarder headquartered in Dubai, with its own full-time employees at peak counted by its finance lead, sits in the second half of the rule. The first draft of its summary said: "Language: English. Basis: headquartered in Dubai; business is conducted in English." That is rung 0 dressed as rung 4.

The note its operations lead completed, and its CEO approved, read differently:

StreamContent typeLanguagesShareBasisStatus
Shipment milestones, 2016-2026Structured codes and short status textenLittle free text to measureRung 3, plus the administrator's descriptionReported
Exception notes, 2016-2026Free texten, aren 85-90%, ar 10-15%Rung 4: 300 notes drawn across all ten years by the operations leadEstimated by the company
Customer email archiveFree text and attachmentsen, arNot countedRung 2Unknown
Driver voice notesAudioar, hi, urNot countedNoneOut of scope: the company would not offer them

The richest stream, the exception notes, sits at or below the 90% line on the company's own count. The summary does not round that up, and it does not conclude the company fails either. It states the range with its basis and leaves the measurement to SourceX. The email archive stays "unknown", with the operations lead named as the person who could count it if the company proceeds.

Labels for estimates, constraints and the unknown

  • Estimates carry three things. The range, the basis (a rung) and the person who produced it. "en 85-90%, operations lead's count of 300 notes, 2026-10-20."
  • Translation is flagged, not absorbed. If a stream was translated, by people or software, write when and how. How translated text is treated is SourceX's decision on the dataset offered.
  • Constraints go in, briefly. Email and call archives carry personal details and may sit under customer contracts; name the constraint and leave the analysis to SourceX's rights review.
  • Unknowns have owners. "Customer email share: unknown; operations lead could count; not yet asked."

Sending the notes and planning the first call

The CEO approves the notes, the relationship owner sends the summary, and the company applies itself through the partner's link. Whoever submits must have signed the SourceX Channel Partner Agreement first. The credit is settled when the company submits a verified application, and it goes to whichever valid introduction came first in the 90-day attribution window.

A useful first call between the company and SourceX:

  1. The company walks through its streams, content types and the languages in each.
  2. It explains how any share was counted: sample size, years covered, counting unit.
  3. It says which streams it would actually consider offering.
  4. SourceX explains what it would measure on that scope.
  5. Unknowns are assigned, with dates, and the company decides on a next step.

For the arithmetic of an English-share inventory across systems, see the English-language eligibility guide; for companies whose records are mostly in another language, whether non-English documentation can be considered.

Where a language note can mislead

  • Self-counts drift upward. A sample from one recent year, or from the team that writes in English, flatters the share. Field 6 and the sample description expose it; they do not cure it.
  • The rule names a share, not a unit. "At least 90% of the data" can be counted in records, words or minutes of audio, and each can give a different answer for the same archive. The note says which unit was used.
  • Code-switching defeats simple counts. A single message can move between two languages; whether it counts as one or the other is a judgment the note should disclose.
  • Programs move. Language preferences differ by program, and a company that does not fit one may fit another later.
  • Tags look precise. A tag records what someone reported. It is not a measurement, and a summary should never imply that it is.

Platform teams setting up the referral side across a portfolio can start from the overview for venture platform teams.

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 every stream in the summary has a language note with a basis the company can stand behind, register for a partner link, compare it with who qualifies and read the program terms before the CEO sees the summary.

Analysts who are looking at SourceX itself as a possible investment should treat that as its own decision, separate from any introduction; the SourceX investor overview handles requests for investor materials.

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 a company in the U.S. need to report an English share at all?

The published rule sets no language share for companies in the U.S., Canada and the EU. State the language of each record stream anyway: SourceX's programs differ by language, and one line in the note saves a question later.

Can the language of a company's website stand in for the language of its records?

No. A website shows which audience the company writes for, which is the second rung of the evidence ladder. It can justify a question to the company, never a share in the note.

Should records that were translated into English be counted as English?

Flag them instead of counting them. Write which streams were translated, when and how, and leave it to SourceX to decide how translated text is treated for the dataset actually offered.

Sources

  1. refer.sourcex.si/terms
  2. refer.sourcex.si/who-qualifies
  3. rfc-editor.org/rfc/rfc5646.html
  4. sso.agc.gov.sg/Act/CONS1963
  5. 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