Do thresholds in mediation analysis matter? A step-by-step methodological framework for quantitative researchers using PLS-SEM

Mediation analysis is increasingly used in behavioural, management, tourism, hospitality, and social science research, particularly with the growing popularity of Partial Least Squares Structural Equation Modelling (PLS-SEM). However, researchers face challenges in evaluating mediation, often overemphasising statistical significance and thresholds while overlooking predictive relevance, theoretical expectations, and substantive significance. This paper aims to systematise methodological guidance for evaluating mediation in PLS-SEM. Through a structured critical review of methodological literature, it examines the assumptions, applications, and limitations of commonly used threshold-based decision rules. Thematic synthesis and critical comparison inform a context-sensitive, step-by-step framework integrating measurement and structural model assessment, predictive power, bootstrap-based significance testing, indirect effects, mediation classification, and effect sizes. The paper advances mediation analysis by integrating previously fragmented assessment criteria into a coherent decision-making framework, clarifying the limitations of mechanical threshold application, and aligning statistical evidence with predictive performance and theoretical reasoning. This contribution provides researchers with a systematic approach to more defensible mediation decisions and offers reviewers and editors clearer grounds for evaluating methodological rigour, interpreting mediation findings, and assessing the transparency of reported results.