A data-driven framework for assessing seller and payment risk in E-commerce marketplaces

E-commerce marketplaces encounter persistent risks from seller fraud and payment defaults, threatening financial stability and trust. This article proposes a robust framework to evaluate these risks by leveraging seller profile data (e.g., business tenure, geographic location, transaction history) and performance indicators (e.g., order completion rates, customer reviews, return frequencies). It employs machine learning techniques, including logistic regression and random forest models, to predict seller reliability and payment risk with precision. A risk scoring system classifies sellers into low, medium, or high-risk tiers, facilitating targeted actions such as intensified scrutiny or real-time monitoring. Applied to a major marketplace, the framework reduced fraudulent transactions, boosted payment recovery, decreased customer complaints, and minimized manual review requirements—all while maintaining low false positive rates. This article contributes to fintech risk management by providing a scalable, data-centric solution for seller oversight. Future enhancements could integrate real-time behavioral tracking, blockchain technology for tamper-proof records, and advanced NLP for textual analysis. It empowers marketplaces to mitigate financial and reputational losses proactively, advancing the field of e-commerce risk assessment.