When defines where: how rural digital divide, customers’ mobility and usage patterns have shaped mobile infrastructures

Expanding and upgrading communication networks in rural areas is particularly challenging due to high investment costs and limited prospects for cost recovery. In this context, researchers and policymakers widely view mobile infrastructures as the most viable solution, given their flexibility and lower deployment costs. However, techno-economic assessments that support this view often rely on population-based demand models, an approach which fails to capture the growing complexity of contemporary usage patterns. This study investigates the key drivers of mobile traffic demand by assessing the predictive value of multiple variables, including Fixed Wireless Access, non-resident mobility, and the spatiotemporal dynamics of service usage. Using voice and broadband traffic from all sites of a major Spanish operator, combined with FWA subscriptions and visitor counts per municipality, we characterize usage patterns across different municipality sizes. Comparing linear models with a Random Forest approach shows that population-based models perform poorly, especially in small municipalities. Overall, the results indicate that network planning should adopt multi-source, non-linear demand models that account for seasonality, visitor flows, and FWA adoption to better inform investment decisions and ensure service quality across diverse geotypes. • Voice and data show distinct patterns across daily, weekly, and yearly scales. • Municipality size strongly shapes telco traffic across all service types. • Population is not a good predictor of demand, especially in small towns. • Traditional techno-economic models fail when checked against real-world data. • Assessments should include mobile visitors, seasonality, and FWA penetration.