Hierarchical phononic materials have sparked interest as they can manipulate the dispersive characteristics of vibrational waves at multiple frequency ranges. Typically, the hierarchical length scales are relatively close, which makes multiscale methods unsuitable. Thus, the design process can only be carried out on a fine grid, which significantly increases the computational cost. In this paper, hierarchical unit-cell templates are presented, where each length scale results in a unique bandgap. The coarse-scale bandgap is mostly unaffected by the introduction of fine-scale features, despite the coarse length scale being only twice that of the fine scale. Thus, a scale-separation effect is imposed between the close length scales, leading to a process where the hierarchical patterns are obtained consecutively, starting at the coarse scale and then refining the fine scale, which reduces the computational cost. The proposed method extends the unit-cell template method (an interpretable machine-learning method specifically created to uncover global patterns in phononic materials) to deal with fabrication constraints and multiple grids and design objectives. The hierarchical patterns are not predefined, unlike existing approaches, e.g., self-similar patterns. The hierarchical templates offer a new framework for extracting effective patterns in hierarchical phononic materials, leading to a flexible and efficient design process.