Client
A mid-sized manufacturing business working with 150-200 regular vendors.They used Tally for accounting, Excel for vendor follow-up, and email or phone calls to manage delivery and payment issues.
The challenge
As the business grew, vendor decisions became harder to manage through spreadsheets alone. Purchase teams had one view of suppliers, accounts had another, and there was no single place to understand which vendors were reliable, which ones were slipping, and where payment risk was building up.
The business wanted more than a scoring script. They needed a working system that could take accounting and operational data, calculate vendor reliability in the backend, and present the results in a way purchase and accounts teams could actually use day to day.
They were also clear that this had to remain practical for a smaller company. The system had to fit around Tally exports and existing processes, not introduce a large procurement platform with heavy licence costs or long implementation cycles.
What we did
We built a complete vendor reliability platform using Python Fast API backend with React frontend, PostgreSQL database, and Docker deployment.
We created scheduled ETL jobs using Python + Pandas to ingest Tally XML/Excel exports daily. The pipeline handled inconsistent vendor names, missing delivery dates, and partial quality data, consolidating 18 months of transactions into a clean PostgreSQL data warehouse with 25+ vendor features.
The core scoring engine used scikit-learn gradient-boosted models trained on delivery delays (40% weight), payment cycle consistency (30%), order volume stability (20%), and quality feedback (10%). New vendors started with conservative baseline scores that matured over 3-6 months of data. Fast API served real-time vendor lookups and bulk score updates with full audit logging.
We built a responsive web application for purchase and accounts teams with:
- Executive dashboard showing vendor risk distribution, recent score changes, and top alerts
- Vendor list with column filters (region, category, reliability band, payment status)
- Vendor detail pages with score trends, delivery performance charts, payment cycle analysis, and recent transactions
- Operational workflows for marking vendors as “Approved,” “Watchlist,” or “Manual Review” with internal notes
The React app used React Query for real-time data sync, Chart.js for performance visualizations across desktop, tablet, and mobile. Role-based access ensured purchase teams saw delivery-focused views while accounts focused on payment risks.
The result
The client moved from fragmented spreadsheet-based vendor tracking to a single working platform for vendor visibility and decision support. Purchase and accounts teams now had a shared view of vendor performance, payment risk, and suppliers needing attention.
The backend gave the business a more structured way to calculate reliability from historical data, while the frontend made those insights usable in everyday workflows. Instead of reacting after delays or issues had already happened, teams could identify risk earlier and review vendors in a more consistent way.
Most importantly, the business got a tailored internal platform that matched its size and operating style. It delivered the benefits of a modern procurement dashboard and Python-based scoring backend without forcing the company into a heavyweight enterprise procurement product.