Trusting Social

A credit score from a phone number, built on anonymized telecom data, is the product, alongside a fraud score from the same data and AI driven onboarding. Its borrowers have no credit file for a rules system to read, so without the models nothing is left to sell.

Lenders make the decisions on scores presented as instant, and AI agents for sales and analytics run with minimal human supervision while aiding staff. Whether scores feed automated approval, what thresholds apply and whether a person reviews adverse outcomes are not described.

Gini, KS, approval rates and a validation method are unpublished, and no reason codes are described for lenders that must explain an adverse decision. Loss figures measure results without explaining how a score was reached.

More than a billion consumers scored and credit loss reductions of 63 percent for credit cards and 48 percent for motorbike financing are published, with no baseline or method. UNO Digital Bank and the Masan Group appear as partners on the blog, but no lender is joined to a figure.

Anonymized and masked data feeds the scores, and no personal information of data subjects is retained, a categorical commitment. Onboarding and fraud data follow a different regime under the Vietnam notice, and whether one lender's outcomes improve models scoring another's applicants is not stated.

Updated November 2024 under Decree 13/2023, the Vietnam privacy notice reaches people processed for credit scoring, eKYC and fraud, covering biometric data, consent with withdrawal, and access, correction and deletion rights. Retention runs only as long as necessary, with no processors identified.

ISO/IEC 27001 certification for the group is announced on the blog. Certificate scope, a trust center and a SOC 2 report are not published.

The GDPR is named, and the blog announces selection by the Reserve Bank of India for a fraud initiative and entry to Visa's Fintech Partner Connect program. Credit information rules governing alternative scoring in Vietnam, Indonesia and India are not named.

Telecom data and phone usage can stand in for income, location and social group, so bias sits inside the credit decision. The inclusion case for borrowers conventional scoring cannot assess has no evaluation behind it, with no fair lending testing, disparate impact analysis or adverse action guidance published.

Someone scored from a phone number by a company they have never dealt with is not told which signals drove the score. Trusting Social describes no route to contest a score once a lender acts on it and publishes no guarantee or indemnity.

Microsoft Azure and OpenAI services are named behind its AI agents. Telecom operators and data partners supplying its signals are not named, so durable access in each market cannot be judged.

Delivery runs from a phone number, and acquisition goes through embedded lending arrangements, but no origination system, bureau or other lender system is named as an integration. No developer documentation is linked.

Scoring is processed on premises under partner data center control, and the Vietnam notice stores data in Vietnam with backups in other regions. That on premises model gets one line, and hosting for India and Indonesia is left out.

Scores, onboarding and customer acquisition products carry no published price or basis of charge, and engagement runs through a contact request. Per query pricing would suit a phone number score, but neither unit nor rate is stated, nor how acquisition arrangements with lenders are charged.

Commercial and digital banks and consumer finance companies, more than 130 institutions across India, Indonesia and Vietnam, use its scoring, fraud and onboarding products, with a digital bank partnership in the Philippines. Material is organized by product rather than institution type.