Although finite-state models have been used in software modeling for quite some time, a general method for building and manipulating such models which directly relates to a program?s structure is not readily available. We fill this gap by constructing a canonical finite-state model M(P) from the flowchart of a program P. Next, we present several methods for simplifying M(P) which correspond to creating higher level models for P and to improving P by eliminating its design flaws. Finally, we show that states based on data-values and their abstractions give us a greater flexibility in creating finite-state models that can be used in practice to build the models from requirements.