Nuclear data are the foundation of simulation and design in the nuclear industry. The success of commercialising thermonuclear fusion will be based on a set of highly accurate simulations used in design, optimisation and safety analyses. This work focuses on the often overlooked, pre-processing stage of nuclear data. The effect of legacy methods in a fusion context is a concern within the community, but has never been quantified. The sensitivity of fusion neutronics to pre-processing was determined using a set of codes and methods developed as part of this thesis. Legacy pre-processing methods demonstrated a difference between the processed and unprocessed distributions of up to 20%. Simple Monte-Carlo radiation transport simulations exhibited sensitivity within energy distributions for small models (< 5 mfp). Alternative data formats did not improve simulation results sufficiently to justify their implementation. Complex, fusion specific models showed a general insensitivity to the pre-processing when run to the current levels of statistical precision. Future recommendations are to process all future data libraries into the cumulative tabulated probability format. Improved methods are not required at this stage as the core data libraries are incomplete and sometimes inaccurate. Only after the libraries have improved will pre-processing become significant.