AI-Based Revenue Leakage Detection Models Using Transaction-Level Financial Data: A Review

Abstract

AI-based revenue leakage detection using transaction-level financial data has gained importance due to increasing pricing complexity, automated billing processes, and high-volume digital transactions. This study quantitatively examined four detection constructs—pricing compliance detection, authorization integrity, temporal anomaly identification, and adjustment behavior monitoring—and evaluated their relationships with overall revenue leakage detection effectiveness. A structured survey design was applied, and responses were collected from 210 participants across transaction-intensive industries, including telecommunications (22.4%), e-commerce/retail (21.0%), healthcare/insurance (18.6%), and financial services (17.1%). Most respondents reported direct involvement in revenue-cycle activities (61.0%), and 75.7% reported intermediate-to-advanced familiarity with analytics and AI tools. Descriptive results indicated consistently positive construct scores, with mean values of 4.12 (SD = 0.61) for pricing compliance detection, 4.08 (SD = 0.65) for authorization integrity, 3.94 (SD = 0.70) for temporal anomaly identification, 3.89 (SD = 0.74) for adjustment behavior monitoring, and 4.05 (SD = 0.63) for leakage detection effectiveness. Reliability analysis confirmed strong internal consistency, with Cronbach’s alpha values ranging from 0.81 to 0.88 across constructs. Multiple regression analysis demonstrated that the predictors jointly explained substantial variance in leakage detection effectiveness (R² = 0.62; adjusted R² = 0.61; F = 83.40; p < .001). Pricing compliance detection produced the strongest standardized effect (β = 0.38; t = 6.52; p < .001), followed by authorization integrity (β = 0.29; t = 5.11; p < .001), adjustment behavior monitoring (β = 0.21; t = 3.88; p < .001), and temporal anomaly identification (β = 0.17; t = 3.09; p = .002). Multicollinearity remained acceptable (VIF = 1.41–1.72). Hypothesis testing supported 4 of 5 hypotheses (80%), with one interaction hypothesis rejected (β = 0.06; p = .118). Overall, the findings demonstrated that effective transaction-level revenue leakage detection was strongly associated with pricing integrity, governance controls, temporal monitoring, and adjustment analytics.