Optimizing New ATM Locations Using the P-Median Model and Geographic Information Systems

Abstract

ATM accessibility remains a spatial planning challenge because existing facility configurations may not adequately reflect heterogeneous patterns of customer demand, population distribution, and economic activity. This study develops an integrated Geographic Information Systems (GIS)–P-Median framework to optimize new ATM locations using multidimensional spatial demand indicators and network-based accessibility. Banking transaction activity, customer distribution, population, economic characteristics, points of interest, and transportation networks were integrated to construct demand weights, screen feasible candidate locations, and generate demand-to-facility network distances. The P-Median model was subsequently solved for alternative expansion levels of one, two, three, and five new ATMs under baseline, moderate-growth (+20%), and high-growth (+50%) demand scenarios. The existing network recorded a Total Weighted Distance of 156.933, an average weighted access distance of 2.853 km, and only 40.58% demand coverage, leaving 59.42% underserved. Across new-facility scenarios, increasing the number of ATMs consistently reduced weighted accessibility distance and increased coverage, although marginal benefits diminished at higher facility counts. The optimal facility sets remained spatially stable across demand-growth scenarios, while Total Weighted Distance increased with demand intensity. The findings demonstrate that GIS-integrated P-Median optimization can identify strategically valuable expansion locations and reveal trade-offs between investment scale, accessibility improvement, and demand growth. The framework provides a reproducible decision-support basis for evidence-based ATM network planning and offers a foundation for future capacity-constrained, integrated existing-plus-new facility optimization and supports more equitable allocation of banking infrastructure across underserved areas.