Traditional credit scoring models like FICO are fundamentally broken for modern small businesses and gig workers. They rely on static, historical credit checks rather than real-time operational health, leaving millions of creditworthy businesses—from SaaS startups with recurring revenue to suppliers waiting on unpaid invoices—unable to secure affordable capital. We were inspired to build Cash-Flow Ledger to democratize access to capital by evaluating businesses based on what actually matters: their real-time cash flow, revenue stability, and operational trajectory.

What it does Cash-Flow Ledger is an AI-driven fintech platform that replaces traditional credit scores with machine learning models trained on bank and POS transaction data. It:

Scores Businesses: Analyzes transaction history to compute a transparent, factor-decomposed cash flow score based on revenue consistency, growth trends, balance stability, and vertical-specific indicators. Routes to Capital: Automatically matches businesses to tailored financial products—such as Revenue-Based Financing, Invoice Factoring, Cash-Flow Term Loans, or Community Marketplaces. Empowers Lenders: Provides an institutional-grade Lender Admin Dashboard with complete transaction histories, risk factor analysis, and immutable audit logging for underwriting decisions. How we built it Frontend: Built with React 19, TypeScript, Vite 8, and styled with Tailwind CSS v4 for a high-performance, dark-themed responsive dashboard. Machine Learning Pipeline: Developed a Python feature-engineering model using scikit-learn algorithms tailored to four verticals (Subscription, B2B Supplier, Retail, and Gig Worker). Backend & Database: Integrated Supabase for PostgreSQL data storage, Row Level Security (RLS) policies for user data isolation, and Deno-based Supabase Edge Functions (ml-score) for server-side scoring execution. Challenges we ran into Vertical-Specific Feature Engineering: Normalizing wildly different transaction signatures—like daily retail POS settlements versus monthly SaaS subscription payouts—into a single, fair scoring system. Row Level Security (RLS) Complexities: Configuring bulletproof database policies in Supabase so that business users can only view their own financial records, while lender admins retain read-all and audit-write permissions. Real-Time Client & Server Synchronization: Building a seamless fallback architecture where the frontend can operate in lightweight mock mode while smoothly transitioning to live Supabase Edge Function execution. Accomplishments that we're proud of Transparent Scoring Decomposition: Instead of returning a black-box number, our engine breaks down exact percentage contributions (Revenue Consistency 30%, Growth Trend 25%, Balance Stability 25%, Vertical Signals 20%). Zero Rejection Design: Designed a routing engine that never flatly rejects a business; if primary products aren’t a fit, businesses are routed to supportive community lender marketplaces to build credit. Complete End-to-End Prototype: Delivered a fully working full-stack application featuring live user onboarding, transaction analysis, automated product matching, and admin underwriting flows. What we learned Transaction Data Beats Static Credit: Granular cash-flow metrics provide a significantly clearer picture of short-term business solvency than personal credit history. Security & Auditing in Fintech: In financial platforms, immutable audit logs and strict data isolation aren't optional features—they are fundamental architectural prerequisites. What's next for Cash-Flow Ledger Plaid & Open Banking Integration: Replace mock bank linking with live Plaid API connections for instant real-time bank statement fetching. Credit Bureau API Sync: Integrate real-time credit reporting API feeds (Experian, TransUnion) to complement cash-flow metrics with traditional background checks. Institutional Lender APIs: Partner directly with lending institutions to enable one-click automated loan application submission and instant payout dispatch.

Built With

Share this project:

Updates