Education has long been a foundational element of societal advancement, with primary and secondary education playing a pivotal role in shaping cognitive and social outcomes. As educational systems face increasing complexity—marked by data abundance, external pressures, and resource constraints—evaluating institutional performance requires scientifically grounded, data-driven methodologies. This study applies a two-stage network data envelopment analysis (NDEA) model to assess the efficiency of 45 high schools in Portugal. The model decomposes school operations into two interconnected stages: resource allocation and educational output generation, with a particular emphasis on mathematical proficiency. The analysis integrates the Anderson–Peterson (AP) super-efficiency method to provide a comprehensive ranking of schools, even among those deemed technically efficient under standard DEA. Findings indicate that only two schools achieved full efficiency across both stages, serving as performance benchmarks. The remaining schools displayed varying degrees of inefficiency, particularly in translating internal processes into strong mathematics outcomes. These insights highlight the importance of targeted interventions at both resource and process levels. The study concludes with practical recommendations for policymakers, including data-informed resource planning, investment in math-specific programs, and institutional benchmarking. The proposed model offers a robust framework for improving educational performance and guiding strategic initiatives in underperforming schools.