Real surfaces often consist of several materials, which may be distributed in intricate patterns. In computer graphics, such composite materials are represented by a 6D function known as the Spatially Varying Bidirectional Reflectance Distribution Function, or SVBRDF. Producing representations of the SVBRDF that are simultaneously efficient and intuitive to work with is an ongoing research challenge. We propose a new algorithm, Nonsmooth Alternating Constrained Least Squares (nsACLS), which combines the best aspects of two existing algorithms for sparse nonnegative matrix factorization. We fuse this algorithm with an interactive GUI and an efficient, distributed implementation. The resulting tool enables users to quickly produce intuitive and realistic spatial factorizations of the SVBRDF.