The minimization of cost, power consumption and time-to-market of DSP applications requires the development of methodologies for the automatic implementation of floating-point algorithms in fixed-point architectures. In this paper a new methodology for evaluating the quality of an implementation through the automatic determination of the Signal to Quantization Noise Ratio (SQNR) is under consideration. The theoretical concepts and the different phases of the methodology are explained. Then, the ability of our approach for computing the SQNR efficiently and its beneficial contribution in the process of data word-length minimization are shown through some examples.