<![CDATA[Student behavior assessment in Indonesian secondary schools typically relies on a rigid point system that tracks compliance but ignores peer-observable social behavior, lacking a mechanism to compensate across dimensions. This study presents SiPandu, a Fuzzy Mamdani-based Decision Support System that integrates three variables, attendance percentage (X1), cumulative behavior points (X2), and peer-assessment scores (X3) into a single adaptive score using the ADDIE R&D model. These variables were standardized to a [0,100] scale and processed via 27 IF-THEN rules with discrete centroid defuzzification, yielding four categories: Needs Guidance, Fair, Good, and Excellent. Content-validated by a school counselor, the peer-assessment instrument aligns with Indonesian Ministerial Regulation No. 10 of 2025. Built with Next.js and Supabase alongside a 7-badge gamification layer, the system was piloted on 39 students at a public senior high school in Bandung. Black-box testing achieved a 100% success rate across 20 scenarios, confirming proper Row Level Security. System Usability Scale (SUS) testing (n=17) yielded a mean score of 65.88 (OK/Marginal), revealing a gap between administrative users (92.5) and students (62.33). Manual simulation confirmed the model's compensation mechanism, where strengths in one dimension offset weaknesses in another without resorting to binary pass/fail outcomes. These findings show that Fuzzy Mamdani inference effectively softens the rigidity of conventional behavior assessments while remaining interpretable for non-technical homeroom teachers.]]>
