Data Cleanroom · Use case
Alternative Credit Scoring
Millions of people have no credit history but years of mobile history. The Data Cleanroom lets lenders turn that telco behaviour into a credit signal — with the customer’s consent, without the telco sharing subscriber data, and without the lender ever holding it.
The Problem
No credit history doesn’t mean no history
First-time borrowers, young people, gig and informal workers, and customers outside big cities often have little or no bureau record. Lenders can’t assess them, so they’re declined or priced as high risk — even when they would repay reliably.
Many of these same people have held the same mobile number for years, top up or pay their bill on a regular rhythm, and use their phone in stable, predictable ways. That behaviour says a lot about stability and ability to pay.
But a telco can’t simply hand subscriber data to a lender, and a lender shouldn’t collect raw telco records just to make a decision. The signal exists — it has just been locked away.
The Signals
Telco signals that predict stability
None of these are shared as raw data. They’re computed inside the cleanroom and combined into a score.
Number tenure
How long the customer has held the same number — long tenure signals stability.
Payment regularity
Whether bills are paid on time, or prepaid top-ups happen on a steady rhythm.
Spend consistency
How stable top-up or bill amounts are over time — a proxy for steady income.
Usage stability
Consistent calling and data use rather than sudden, erratic changes.
Location stability
A consistent home and work area over time — measured as a pattern, never as tracking.
Device stability
How often the customer changes handsets or SIMs.
How It Works
How the Data Cleanroom makes it work
Customer consents
The applicant agrees, in the lender’s app or branch, to a telco-based assessment.
Lender sends an encrypted query
The lender submits the applicant’s encrypted identifier to the cleanroom — no raw personal data changes hands.
Score computed privately
The telco’s behavioural features and the agreed scoring model are combined using multi-party computation on encrypted inputs.
Only the score comes back
The lender receives a score or risk band with reason codes, and uses it alongside its own policy and any bureau data.
The Outcome
What this changes
Who It's For
Who it’s for
Banks & NBFCs
Extend credit to new-to-credit customers with a signal beyond the bureau.
Fintech & microfinance lenders
Underwrite small-ticket and first-time loans faster, with less guesswork.
Telecom operators
Turn network behaviour into a credit signal while subscriber data stays in-house.
Privacy
Private by design
Consent first
Assessments run only with the applicant’s explicit agreement.
Data stays at its source
Telco records never leave the telco; the lender never holds them.
Encrypted end to end
Identifiers and features stay encrypted throughout the computation.
Only a score is released
The output is limited to the score and reason codes — never the underlying behaviour.