Automotive seats are commonly described by one-dimensional measurements, including those documented in SAE J2732. However, 1-D measurements provide minimal information on seat shape. The goal of this work was to develop a statistical framework to analyze and model the surface shapes of seats by using techniques similar to those that have been used for modeling human body shapes. The 3-D contour of twelve driver seats of a pickup truck and sedans were scanned and aligned, and 408 landmarks were identified using a semi-automatic process. A template mesh of 18,306 vertices was morphed to match the scan at the landmark positions, and the remaining nodes were automatically adjusted to match the scanned surface. A principal component (PC) analysis was performed on the resulting homologous meshes. Each seat was uniquely represented by a set of PC scores; 10 PC scores explained 95% of the total variance. This new shape description has many applications. For instance, seats can be classified and grouped, and the distinctive features and dimensions that characterize each group can be identified. Seats can be compared by evaluating a distance metric in PC space. Overall, seat design and benchmarking can benefit from the statistical insight offered by this new approach.