We report on the performance evaluation of greedy parsing with a single step lookahead (which we call flexible Parsing or FP as an alternative to the commonly used greedy parsing (with no-lookaheads) scheme. Greedy parsing is the basis of most popular compression programs including UNIX compress and gzip , however it usually results in far from optimal parsing/compression with regard to the dictionary construction scheme in use. Flexible parsing, however, is optimal [MS99], i.e. partitions any given input to the smallest number of phrases possible, for dictionary construction schemes which satisfy the prefix property throughout their execution. We focus on the application of FP in the context of the LZW variant of the Lempel-Ziv'78 dictionary construction method [Wel84, ZL78], which is of considerable practical interest. We implement two compression algorithms which use (1) FP with LZW dictionary (LZW- FP ), and (2) FP with an alternative flexible dictionary (FPA as introduced in [Hor95]). Our implementations are based on novel on-line data structures enabling us to use linear time and space. We test our implementations on a collection of input sequences which includes textual files, DNA sequences, medical images, and pseudorandom binary files, and compare our results with two of the most popular compression programs UNIX compress and gzip . Our results demonstrate that flexible parsing is especially useful for non-textual data, on which it improves over the compression rates of compress and gzip by up to 20% and 35%, respectively.