By default, metrics are calculated at the lowest attribute level (most granular data) to ensure accuracy. Calculating at higher levels first can lose precision due to aggregation. The lowest level provides the most detailed foundation for metric calculations.