Credit Scoring

Uses alternative data and machine learning to score borrowers, especially where traditional credit history is thin or unavailable.

Business impact

• repayment rate — Tala reports repayment rates above 90% on loans that average $50.
• customers scored — Bank BRI says it has scored 75.5 million customers.
• approval rate — Worth says customers using its platform have increased approval rates by 37%+.
• time to decision — The evidence describes instant or near-instant credit decisions in Tala and Bank BRI.

Data requirements

• mobile phone data (text) — Tala uses phone usage and communication-related signals.
• app activity (text) — Tala tracks behavior through its lending app to support scoring.
• social media channels (text) — Tala’s article says social media activity is among the alternative signals used.
• ID scans (image) — Bank BRI customers scan their ID in the app as part of the scoring flow.
• first- and third-party authoritative data (tabular) — Worth combines multiple data sources to resolve business identities and score SMBs.
• alternative financial data (tabular) — NinjaHoldings uses alternative financial data with machine learning models.

AI methods and techniques

• machine learning — Used to infer creditworthiness from alternative and first-party data.
• alternative data scoring — Uses nontraditional signals such as mobile phone data, app activity, and other daily-life indicators to estimate repayment risk.
• big data analytics — Used in the bank examples to score customers at scale and support lending decisions.
• proprietary AI/ML models — Used in Worth to produce its unified credit score and support underwriting and monitoring.

Industries

Real-world evidence

5 documented case studies on record.

Companies using this: Tala, Bank Rakyat Indonesia (Bank BRI), NinjaHoldings, Worth AI, Affirm.