Skip to content

Repository files navigation

MerchantFloat

Working capital loans for Nigerian merchants — scored from Interswitch POS transaction data.

Live demo: https://merchantfloat.vercel.app GitHub: https://github.com/David-Pjs/merchantFloat


The Problem

41 million informal merchants in Nigeria process ₦18 trillion annually through POS terminals — yet 84% have zero credit bureau profile. Banks want to lend but have no risk data. The data exists inside Interswitch. Nobody was reading it.

The Solution

MerchantFloat turns Interswitch POS transaction history into a credit score. Merchants enter their BVN, we scan 90 days of Quickteller transaction data, generate a score across 6 signals, and return a loan offer in under 60 seconds. Repayment auto-deducts 10% from each daily Quickteller settlement — no phone calls, no guarantors, no collateral.

Interswitch APIs Used

  • Quickteller Payment API — virtual account generation, card payment processing, transaction verification
  • Interswitch Passport API — BVN identity verification

Credit Scoring Model (6 signals, 100 pts)

Signal Max What it measures
Revenue Volume 25 Monthly naira processed
Trading Consistency 25 Days active out of 30
Revenue Stability 10 Weekly variance (coefficient of variation)
Refund Rate 20 Returns as % of total transactions
Settlement Speed 10 Avg hours for Quickteller to clear funds
Account Coverage 10 Number of linked bank accounts

Score ≥ 50 = eligible. Loan = 15–20% of monthly revenue, scaled by score.

Key Features

  • BVN verification via Interswitch Passport API (one-tap demo flow)
  • Live Quickteller payment with real Wema Bank virtual account
  • Score simulator — merchants see exactly what to fix to qualify
  • Credit ladder — shows loan limits after each repayment cycle
  • Plain English score explanation per merchant
  • Auto-deduction repayment model (no contacts called, ever)
  • Loan-to-revenue ratio displayed on offer card

Demo Merchants

Merchant Score Outcome
Emeka Nwosu (Cold Room) 85/100 ₦750,000 approved
Hajiya Ramatu (Agriculture) 62/100 ₦320,000 approved
Tunde Adeyemi (Electronics) 29/100 Declined — high refund rate

Tech Stack

  • Next.js 16 + TypeScript
  • Tailwind CSS v4
  • Recharts (score visualisation)
  • FastAPI (AI scoring engine)
  • scikit-learn (K-Means clustering, credit model)
  • Vercel (deployment)
  • Interswitch Quickteller + Passport APIs

Team Contributions

David Uhumagho — Team Lead / Full-Stack Developer

  • Product concept, research, and market validation
  • Full frontend implementation (all pages: onboarding, dashboard, loan flow, repayment, checkout)
  • Backend API routes: score engine, Quickteller integration, BVN verification, webhooks
  • Interswitch OAuth2 token management and transaction verification
  • Credit scoring model design and calibration
  • Merchant data research and synthetic dataset creation from real POS transaction patterns
  • Deployment to Vercel with production environment configuration
  • Demo story and pitch narrative

Onwukamuche Onyinyechi Lynda — Data Scientist / AI & ML Engineer

  • Designed and generated a synthetic dataset simulating real POS transaction behavior over 180 days per merchant
  • Engineered key features: daily revenue, transaction count, refund count, returning customer ratio, avg settlement delay, peak sales hour, business category
  • Built diverse merchant profiles: high-performing, medium-performing, low-performing, and volatile merchants
  • Implemented feature engineering pipeline aggregating transaction-level data into merchant behavioral profiles
  • Applied K-Means clustering (n_clusters=4) with StandardScaler normalization for merchant segmentation
  • Developed rule-enhanced segmentation: Elite (High Growth), At Risk (Settlement Delay), Volatile (Unstable), Steady (Reliable)
  • Designed custom credit scoring system (300–850 scale): customer loyalty (25%), revenue consistency (25%), transaction activity (15%), revenue strength (35%)
  • Implemented risk classification: Low Risk (≥700), Medium Risk (520–699), High Risk (<520)
  • Built dynamic loan offer logic: low risk → up to 15 days revenue, medium risk → up to 7 days revenue
  • Developed AI insights engine providing actionable merchant recommendations (e.g. reduce settlement delays, improve retention)
  • Built and validated standalone AI scoring pipeline — confirmed producing live results across merchant segments

Selimat Akinwale — Product Manager

  • Delivered the full Product Requirements Document (PRD) for MerchantFloat
  • Defined the problem and validated the idea against market data
  • Prioritised features against the 3-day buildathon deadline
  • Wrote user journey maps and flow documentation
  • Conducted competitor analysis across Nigerian and global fintech lending products
  • Stakeholder mapping and business model definition
  • Built the risk register and SMART goals framework
  • Wrote the full pitch script for Demo Day presentation

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages