Vliv úrovně granularity mediálních dat na kvalitu a výkonnost modelů marketingového mixu

This bachelor’s thesis examines the effect of media data granularity on the quality and performance of marketing mix models. The thesis aims to analyze how the aggregation of media channels affects Marketing Mix Modeling (MMM) results from both statistical and business perspectives, and to assess whether lower data granularity can be a reasonable compromise between model quality and practical business usability. The thesis first defines the principles of MMM, the specifics of the Robyn package, and the theoretical implications of aggregating media variables. The empirical part then compares models created in Robyn across five time series, each at three levels of granularity: the detailed level of digital media channels, the platform-level aggregation, and the overall digital aggregation. The models are evaluated based on predictive performance, estimated media effects, calculated efficiency, and the interpretability of transformation curves. The results show that aggregation does not necessarily reduce the predictive power of the model and, in some cases, aggregated models achieve better results on test data than the detailed model. From a business perspective, however, its usefulness is limited. Aggregated models often lead to a different interpretation of media efficiency, transformation curves, and therefore also conclusions for media budget allocation. At the same time, the results do not indicate a simple relationship between the distribution of media expenditures and the extent to which the aggregated model preserves the conclusions of the detailed specification.