Writing high performance software requires the programmer to take advantage of multi-core processing. This can be done through tools like OpenMP, which allow the programmer to mark parallel loops. Identifying parallelizable loops, however, is a non-trivial task. Furthermore, transformations can be applied to a loop nest to expose parallelism. Polyhedral compilation has become an increasingly popular technique for exposing parallelism in computationally intensive loop nests. These techniques can simultaneously optimize for a number of performance parameters (i.e. parallelism, locality, etc). This is typically done using a cost model designed to give good performance in the general case. For some problems, the compiler may miss optimization opportunities or even produce a transformation that leads to worse performance. In these cases, the user has little recourse, since there are few options for the user to affect the transformation decisions. In this paper we present PUMA-V, a visualization interface that helps the user understand and affect the transformations made by R-Stream, an industrial strength optimizing compiler based on the polyhedral model. This tool visualizes performance heuristics and runtime performance statistics to help the user identify missed optimization opportunities. Changes to the transformed code can be made by directly manipulating the visualizations. We show an example where performance is greatly improved over the polyhedral model alone by using our tool.