Purpose: Perception‑based surveys are common in forensic accounting research but are often criticised for lacking methodological rigor, particularly regarding common method bias, multicollinearity, and discriminant validity. This study addresses these criticisms by presenting a cross‑sectional survey of 120 finance and audit professionals that implements and transparently reports a comprehensive suite of diagnostic checks – including Harman’s single‑factor test, full collinearity VIF, confirmatory factor analysis (CFA), average variance extracted (AVE), composite reliability (CR), and the HTMT criterion for discriminant validity. The aim is to demonstrate that with rigorous diagnostics, perception‑based studies can produce credible findings that approach predictive validity. Design/methodology/approach: A quantitative cross‑sectional survey was administered to finance, audit, and risk professionals in Nigeria across banking, public sector, manufacturing, and professional services. The questionnaire measured data‑driven forensic analytics (DDFA), internal control systems (ICS), continuous auditing (CA), and fraud detection effectiveness (FDE). Data were analysed using SPSS and AMOS, with full diagnostic reporting: reliability (Cronbach’s α), CFA (factor loadings, AVE, CR), discriminant validity (Fornell‑Larcker and HTMT), multicollinearity (VIF, tolerance), common method bias (Harman’s single‑factor test, full collinearity VIF), and regression assumptions (normality, homoscedasticity, independence, linearity). Hierarchical multiple regression was employed to test associations. Findings: All three predictors are positively and significantly associated with FDE: DDFA (β=0.41, p<0.001), ICS (β=0.36, p<0.001), CA (β=0.30, p<0.001). The model explains 65% of the variance (R²=0.65). Diagnostic checks confirm: Cronbach’s α > 0.80; all factor loadings > 0.60; AVE > 0.50; CR > 0.80; HTMT values < 0.85; VIF < 3.0; Harman’s first factor = 28.5%; full collinearity VIF < 3.3; residuals normal (Shapiro‑Wilk p=0.23); homoscedastic (Breusch‑Pagan p=0.64); Durbin‑Watson = 1.89; Ramsey RESET p=0.27. These diagnostics collectively support the validity of the findings. Practical implications: For researchers, this study provides a template for transparent diagnostic reporting that can elevate the credibility of perception‑based fraud research. For practitioners, the results confirm that analytics, controls, and continuous auditing are each associated with fraud detection effectiveness, with analytics being the strongest. Originality/value: This study contributes a methodological exemplar for the field, demonstrating that “perception” need not imply “weak evidence”. By subjecting data to the full battery of recommended diagnostic tests and reporting all results transparently, it shows how cross‑sectional surveys can approach the rigor required for predictive claims.
