CredoLab
CredoLab Technical Briefing
Published March 2, 2021
Supported by Shopify and Brankas
Overview
Credolab is a Singapore-headquartered company building smartphone-based behavioral credit scoring for the financially underserved. The demo shows how Credolab\'s lender-embedded SDK (Android) captures behavioral metadata alongside the borrower\'s consent during a standard loan-application flow — with no PII ever leaving the phone (50,000 behavioral data points compressed into a 50 KB JSON). The borrower sees transparent Google-policy permission prompts (allow/deny per permission), Credolab reports coverage completeness (e.g. 83% if some permissions are denied), and the score plus metadata (record ID, device ID, timestamp, permissions) surfaces in the lender\'s dashboard in real time. The Credolab score focuses on intent to repay (behavioral patterns matched against known defaulter features via proprietary ML) rather than ability to repay — so it complements existing bureau data rather than replacing it. Use cases span emerging markets without credit bureaus (thin-file approvals on fair terms), developed markets (risk-based / relationship-based pricing), and instant-decision products like short-term loans, buy-now-pay-later, e-commerce, and e-wallets. Credolab has 86 clients across 26 countries.
0:00 Introduction to Credolab
Credolab is a Singapore-headquartered company that builds a behavioral credit-scoring framework from smartphone usage patterns to predict the probability of default — designed to serve the financially excluded and underserved users that traditional credit bureaus can't reach.
1:30 Lender SDK — embedded in the loan app
Credolab is integrated by embedding the lender's SDK into their own mobile app, so the user experience stays seamless. When the borrower applies for a loan they simultaneously consent to Credolab collecting behavioral metadata for scoring.
2:30 Demo app with bank code and unique record ID
Because Credolab never touches PII, each applicant is identified only by a unique invite code known to the lender — no national ID, phone number, or name. A bank-code field routes the record to the correct lender tenant.
3:00 Transparent consent flow and Google permissions
Data privacy is a core pillar: the app explains what permissions are needed, why, and how. Android permission prompts follow Google's policy (explicitly allow or deny) and Credolab reports how complete the dataset is — 83% accessible if the user denies some permissions.
4:30 Score delivered to the lender dashboard in real time
On submission, the score and metadata (record ID, device ID, device type, permissions granted, timestamp) appear in the lender's Credolab dashboard in real time, ready to be fed into their own decisioning framework.
5:30 Intent to repay vs. ability to repay
The Credolab score is designed to complement — not replace — existing credit frameworks. Because it's built on behavioral data, it captures the borrower's intent to repay rather than their income or ability, making it complementary to traditional bureau data.
6:30 Use cases — emerging markets + risk-based pricing
In emerging economies without credit bureaus, Credolab opens credit to thin-file borrowers. In developed markets, it powers risk-based/relationship-based pricing — so lenders offer better rates to borrowers who show strong intent alongside strong ability.
7:30 Privacy architecture — 50,000 data points → 50 KB JSON
Credolab collects ~50,000 behavioral data points from the smartphone but outputs only a 50 KB JSON file — no actual content ever leaves the device. Metadata is processed on Azure and converted to millions of ML features for probability-of-default modeling.
9:00 Best-fit products and global footprint
The score fits instant-decision products — short-term loans, buy now pay later, e-commerce credit, e-wallets — rather than mortgages. Credolab serves 86 clients across 26 countries.
Topics: Identity & KYC, Compliance & Regulation, Lending
