Rapid development of acquisition devices and democratization of easy-to-use modeling tools have resulted in a flood of 3D models. Such data provides a unique opportunity to discover and understand variability in shapes, both inside and across different families (i.e., related shapes). Factorizing such intrinsic variations from acquisition/modeling artifacts can eventually allow us to quantify essence of shape families. Our group has been investigating computational strategies to perform such analysis on large datasets and use the extracted information for shape synthesis and functional modeling. I will present example applications in computational design, re-factoring model databases, and also one example of measuring growth events in plants.