Duplicate invoice payments pose a significant control risk in accounts payable, yet empirical evidence on the relative contributions of staff competence and system alerts to detection accuracy remains limited. This parallel-group randomized experiment (N = 120; n = 60 control, n = 60 alerts; 15 wholesale distribution firms in Semarang, Indonesia) examined whether invoice verification competence and randomized alert access predicted accurate duplicate detection. Participants completed a 20-item competence assessment (KR-20 = .764) and classified 40 invoice candidates (16 duplicates, 24 legitimate) in a standardized interface; the alert system operated at 75% sensitivity and 83.3% specificity. The primary analysis used mixed logistic regression with crossed random effects and Holm correction. Competence was significantly associated with classification accuracy (OR = 1.346 per 10 points, 95% CI [1.284, 1.411], p < .001), confirming H1. Alert access did not significantly improve overall accuracy (OR = 1.130, 95% CI [0.951, 1.342], p = .164), not confirming H2; however, alerts significantly increased duplicate detection sensitivity (OR = 1.548, 95% CI [1.189, 2.017], p = .001) without altering specificity for legitimate invoices. No competence × alert interaction was detected. These findings underscore the primacy of workforce competence in invoice verification and highlight the need to evaluate alert systems against both duplicate detection and legitimate invoice acceptance outcomes.
Copyrights © 2026