Retail Fraud Analysis Retail businesses need

Retail Fraud Analysis

Retail businesses need to detect fraud without unnecessarily flagging legitimate transactions. I built this project using 100,000 retail transactions to explore fraud patterns and identify where financial risk was concentrated.

During data exploration, I found that the original fraud-risk score was stuck at 0.0 across all records, while two location fields contained duplicate information. Rather than inventing a score, I created transparent Low, Medium, and High risk classifications using available security indicators, including transaction velocity, proxy/VPN usage, and high-risk devices.

I used SQL to clean and prepare the data, then built an interactive Power BI dashboard covering the overall picture, fraud detection, and recommendations.

One key finding was that medium-risk transactions represented approximately $7.1 million in financial exposure, compared with $3.3 million for high-risk transactions. The analysis also showed that fraud activity was concentrated in lower transaction-value brackets.

The overview below shows the project, while the full analysis is available on GitHub.