How should a company inventory language coverage for the applicable 90%-English sourcing criterion?
System by system, at metadata level, with counts and sampled shares the company produces itself. The partner's job is to ask for the summary, not to see the records.
First, check whether the criterion applies
The published geography reads: 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 90% of the data is in English.
So the inventory matters for a company in, for example, India, Singapore, the UAE or Brazil. A company in Germany or Quebec does not need it to qualify on location, though language can still affect which datasets fit. The qualitative guide to English documentation covers quick signals; this page is the quantitative worksheet that follows.
The five columns of the inventory
| Column | What goes in it | Who fills it |
|---|---|---|
| System | A plain-text name such as the helpdesk, CRM or document library | Company |
| Record type | Tickets, call notes, transcripts, SOPs, project files | Company |
| Records in scope | Count of records that could be licensed | Company |
| English share | Sampled percentage of records whose main language is English | Company |
| English records | Records in scope multiplied by English share | Company, checked by SourceX |
Use ISO 639 language codes when recording what the sample found ("eng", "hin", "ara"), so mixed results are recorded the same way across systems. The Library of Congress maintains the ISO 639-2 code list as its registration authority. A company can sample with a detection service: Google Cloud Translation's v3 detection, for example, returns a language code with a confidence score between 0 and 1. Google's older v2 method marks its confidence fields as deprecated and advises against thresholds based on them, so a company using it should spot-check results by hand.
A worked inventory: 92.4% or 95.9%?
Illustrative only. A fictional 300-person customer operations company in India runs support for international clients. Its operations lead samples a few hundred records per system and reports these estimates. None of this is SourceX data.
| System | Record type | Records in scope | English share | English records |
|---|---|---|---|---|
| Helpdesk | Resolved tickets | 400,000 | 97% | 388,000 |
| CRM | Account notes | 120,000 | 92% | 110,400 |
| Call platform | Call transcripts | 80,000 | 70% | 56,000 |
| Document library | SOPs and knowledge base | 2,000 | 100% | 2,000 |
| All four systems | Mixed | 602,000 | 92.4% | 556,400 |
| Without transcripts | Mixed | 522,000 | 95.9% | 500,400 |
Both totals clear 90%, but the margin differs, and the reason is visible: the transcripts mix English with Hindi, pulling the overall share down. If the transcripts are not part of what the company would license, leaving them out is accurate scoping. If they are part of it, the 92.4% figure is the honest one. Choosing scope to pass the test, rather than to describe the offer, would only be undone at review.
Three ways the arithmetic can mislead
- Unequal record sizes. One SOP can hold more text than a thousand ticket lines. Counting records weights them equally; counting words would not. Note in the inventory where long and short records sit together.
- Mixed-language records. Decide one rule (main language by majority of text, or a separate "mixed" tally) and apply it to every system.
- Thin samples. A share from twenty records is a guess. Sample randomly across years and teams, and say how many records were checked.
The terms state the 90% threshold but not the unit of measurement, so treat the inventory as the company's evidence for review, not a final ruling. SourceX confirms how the criterion applies to the records actually proposed.
What else the partner should confirm
Language is one test among several. The company also needs 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license the records (client-owned records in an outsourcing business often are not the company's to license), and an authorized contact. The BPO screening worksheet addresses the client-ownership question directly; the manufacturing and engineering and design worksheets cover other sectors. Partners outside the company's country can read the guide for global partners, and the company fit checker runs the basics in a minute.
Keep every question at metadata level: which systems, how many records, which years, what share is English, who owns them, who can authorize a licensing discussion. Never ask for samples, exports or screenshots. What stays unknown until SourceX reviews the company: whether the records match current dataset demand, the rights review result, the final language measure and any offer.
The company's compensation and the referral reward
The company is paid its Actual Supplier Payout under its own agreement; SourceX funds the partner's reward separately, so the reward never reduces the company's payment. Illustrative only: an Actual Supplier Payout of $120,000 would mean a reward of $30,000, and $400,000 would reach the $100,000 per-company cap. Full calculation and status rules are on the rewards page.
Next step
Refer a company. Earn up to $100,000. Default reward: 25% of Actual Supplier Payout, capped at $100,000 cumulatively per referred company. Eligibility and payment conditions apply, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If the company's inventory clears the threshold for the records it would license, register as a partner, confirm the remaining criteria on who qualifies, and keep the geography wording from the program terms on hand when you explain it to the owner.