Refactoring the application will be a more relaxed process if we have some large sets of sample input data with corresponding expected outputs. This is not an easy task because it either brings a database into the mix, or would require abstracting the data access so that a test could use a file-based source.