Inspiration
India has 63 million MSMEs — the backbone of the economy — yet over ₹25 trillion in credit demand goes unmet every year. The problem isn't that banks don't want to lend. It's that most small business owners have no formal credit identity. A kirana store owner in Nagpur processing hundreds of UPI transactions daily is, on paper, invisible to the financial system.We kept coming back to one uncomfortable truth: the data already exists. GST returns, UPI histories, bank statements — it's all there. What's missing is the infrastructure to translate that data into trust. That's what inspired CredSetu — a credit bridge (setu = bridge in Hindi) between India's real economy and formal finance.
What it does
CredSetu is an AI-powered MSME credit infrastructure platform with three core flows:For MSMEs:
Onboard by submitting GSTIN, bank statements, GSTR-3B returns, and UPI transaction history Receive a 0–850 credit score across 6 dimensions: Revenue Growth, GST Compliance, Cashflow Stability, Debt Service Coverage, UPI Payment Velocity, and Fraud Risk Get matched with lenders (banks, NBFCs, fintechs) whose risk appetite fits the business profile Apply to multiple lenders with a single consent-first data share Discount outstanding invoices via TReDS-style invoice financing For Lenders:
Access a clean pipeline dashboard with filterable applications by credit tier, score, and status Update application statuses (Under Review → Approved / Rejected) in real time View portfolio analytics — score distribution, approval rates, pipeline value The platform covers three distinct borrower personas: a kirana retailer (score 682, YELLOW), a textile manufacturer (score 741, GREEN), and an electronics trader with fraud flags (score 541, ORANGE) — each demonstrating how the system handles different risk profiles.
How we built it
We built CredSetu entirely in React, using Vite as the build tool and Recharts for all data visualisation. The full project is split across a deliberate file structure:
tokens.js — a single source of truth for the design system (colours, spacing logic) styles.js — a CSS-in-JS global stylesheet injected via a tag, giving us full control without any external CSS framework data/mockData.js — three fully fleshed-out MSME profiles with 12 months of revenue, cashflow, fraud flags, and application history utils/helpers.js — pure functions for score colour mapping and status colouring components/ — shared UI atoms: TierBadge, StatusPill, ScoreGauge, FactorBars, CustomTooltip pages/ — four top-level pages: Landing, Onboarding, MSME Dashboard, Lender Portal tabs/ — five dashboard tabs broken out as independent components for Overview, Score Breakdown, Marketplace, Invoice Discounting, and Applications</p> <p>No backend, no database — every interaction is simulated with useState and realistic mock data, keeping the demo fast and fully self-contained.</p> <h2 id="challenges-we-ran-into">Challenges we ran into</h2> <p>Designing for two opposite users in one app. The MSME and the lender have completely different mental models. An MSME wants reassurance and clarity; a lender wants filterable data and risk signals. Building a single codebase that served both without either UI feeling compromised required deliberate context switching — and more than a few redesigns of the sidebar and navigation flow. Making credit scores feel human. A number between 0 and 850 means nothing without context. Getting the score gauge, tier badges, factor breakdowns, and improvement tips to work together — so a user understands their score rather than just sees it — took a lot of iteration on the visual hierarchy. Consent-first data sharing UX. In India's AA (Account Aggregator) framework, consent is legally meaningful. We wanted the consent modal to feel real and trustworthy, not like a dismissable GDPR banner. Striking the right tone — specific, time-limited, purpose-bound — in a small modal was harder than expected. Keeping mock data honest. We wanted the three personas to feel like real businesses, not toy examples. That meant crafting revenue curves that told a story (Priya's steady growth vs. Suresh's volatile spikes), fraud flags with plausible explanations, and cashflow ratios that matched the stated scores.</p> <h2 id="accomplishments-that-were-proud-of">Accomplishments that we're proud of</h2> <p>Built a fully functional, multi-persona fintech product in a single hackathon sprint Designed a credit scoring UI that explains why — not just what — making financial data accessible to first-time borrowers The fraud detection flags and score improvement tips make the platform genuinely educational, not just a credit check tool A clean, production-quality codebase split across 16 files with zero hacks — something we could actually hand to a developer and continue building</p> <h2 id="what-we-learned">What we learned</h2> <p>Financial inclusion is a UX problem as much as a data problem. The data to assess MSME creditworthiness already exists in India's GST and UPI ecosystems. The gap is translating that data into something a business owner can understand and act on — and something a lender can trust. Designing for low-trust environments is hard. MSMEs applying for credit for the first time are often anxious and skeptical. Every UI decision — from the consent modal copy to the colour of a tier badge — either builds or erodes that trust. Recharts is powerful but opinionated. Getting custom tooltips, gradient fills, and multi-series bar charts to behave exactly as designed required digging well beneath the surface of the defaults. Scope discipline wins hackathons. We cut invoice discounting down to a simple two-invoice demo rather than a full table, and cut the lender analytics to two charts rather than six. That focus is what let us ship something polished rather than something ambitious but broken.</p>
Built With
- hooks
- javascript
- react
- typescript
- vite
Log in or sign up for Devpost to join the conversation.