Procurement Data Quality · Fuzzy Matching

Supplier Normalization Workbench

A supplier data-quality workbench for standardizing messy vendor names, detecting duplicate supplier records, and creating cleaner supplier-family mappings for spend analytics.

Open Live Demo

PythonStreamlitProcurementData QualityRapidFuzz

Problem

ERP and vendor-master data often contains inconsistent supplier names such as IBM, I.B.M. Corp, IBM Corporation, and International Business Machines. These inconsistencies distort supplier counts, spend concentration, category fragmentation, and sourcing opportunity analysis.

Solution

I built a standalone Streamlit workbench that cleans supplier names, applies known alias rules, uses RapidFuzz fuzzy matching, groups similar suppliers into normalized supplier families, flags risky matches for human review, and exports cleaner supplier data for downstream spend analytics.

Core Features

Supplier name cleaning and standardization

Known alias matching

RapidFuzz fuzzy duplicate detection

Match confidence scoring

False-positive risk flagging

Human review queue

Golden supplier record recommendations

Before-and-after supplier count impact

Exportable normalized supplier data

What It Demonstrates

This project demonstrates procurement data-quality thinking, fuzzy matching logic, human-in-the-loop review design, and the foundational role of supplier normalization in spend analytics and sourcing diagnostics.