Statistical analysis of TRMM-PR column reflectivity profiles is often limited by the fact that, at a given temperature, the same reflectivity may occur in very different convective or stratiform conditions. Vertical profile information (and knowledge of related physical processes) is lost in ensemble analyses, at best only crudely retrievable from Level 2 data by the 2A23 convective/stratiform and bright band classification. This study utilizes traditional cluster analysis to identify natural archetypes (clusters) of similar vertical profiles. Using a 4 million column training dataset, and concurrent NCEP reanalysis data to assign temperature levels to altitudes, "familiar" profile types (clusters) such as shallow stratocumulus, convective warm rain, stratiform-with-bright band, deep convective and overhanging anvil are readily identified from the TRMM vertical column data themselves. 2A23 convective/stratiform and bright band classifications are also used as low-weighted inputs to "nudge" the clustering results. The remaining 70 million columns participating in warm-season precipitation features in the mission's first 3 years are subsequently assigned cluster types based on these cluster definitions. The spatial and seasonal distribution of cluster types provides a unique and physically-based tool to quantitatively identify local "convective regimes". The distribution of assigned profile types within storms can also be used as inputs to a secondary cluster analysis which classifies archetypal storm types, or as inputs to statistical models which, e.g., predict lightning occurrence based on storm vertical structure.