The Paradox of Stability: Bayesian Network Analytics on IoT-Driven Supply Chain Risk Asymmetry in Startup Unmanned Retail

Abstract

 This study addresses the critical challenge of supply chain risk management in startup unmanned retail enterprises, where conventional models struggle to capture dynamic risk transmissions under data scarcity. By innovatively integrating Bayesian network (BN) methodology with expert knowledge and fragmented operational data, we develop a tripartite risk framework encompassing technological, operational, and consumer behavior dimensions. Empirical validation through mixed-methods analysis of Chinese unmanned retail startups reveals three pivotal findings: First, procurement management (X3), network failures (X6), and sensor reliability (X7) emerge as core risk nodes, demonstrating heterogeneous effects across risk levels—network stability paradoxically serves both as a safety foundation (P(T=0|X6=0)=26.82%) and crisis accelerator (P(T=1|X6=1)=52.96%). Second, technology-operations coupling effects amplify risk propagation, where IoT component failures trigger cascading decision biases (e.g., 55.01% medium-risk probability from X3 disruptions). Third, posterior probability analysis identifies payment system stability (X8) as the global leverage point, with 40.42% influence on low-risk states. Theoretically, we extend complex adaptive systems theory by quantifying risk factor asymmetry and validating Haimes' risk coupling dynamics in digital retail contexts. Methodological limitations regarding static probability assumptions and startup sample bias are addressed through proposed dynamic Bayesian network extensions and AI ethics integration in future research.