[en] Analysis of sensitivity profiles and covariance data provides a robust mathematical approach towards assessing the similarity of experimental benchmarks that can be used for validation of criticality calculations. This approach is all the more valuable when dealing with fissile materials for which only a small number of experimental benchmarks exist, since it makes it possible to select other relevant experiments performed with different fissile materials. A practical example of Framatome industrial application is performed on TRIGA fuel, used worldwide in research nuclear reactors. The fissile material of such fuel elements is a metallic alloy of UZrHx intrinsically moderated with an atomic ratio of H/Zr close to 1.6 and a U content ranging between 8 and 47 wt.%, with 20 wt.% enrichment. The experimental validation of such media is typically based on a very small number of ICSBEP (International Criticality Safety Benchmark Evaluation Project) experimental configurations (mainly IEU-COMP-THERM-003). However, the analysis of sensitivity profiles, as well as the contribution of each reaction to prior uncertainty using various covariance data helps pointing out other ICSBEP benchmarks of interest. These experiments can then be used to extend the experimental validation database and justify the bias due to nuclear data and calculation scheme for criticality calculations performed with the French criticality safety code package CRISTAL. Those calculations, currently performed with the TSUNAMI/TSURFER sensitivity/uncertainty sequence of the SCALE package, show the potential industrial interest in these approaches. They also highlight the need to develop in-house expertise in the domain, as well as tools more consistent with CRISTAL package such as MACSENS, currently under development at IRSN. This paper shows an industrial practical application of sensitivity and similarity techniques, namely the extension of the validation database for UZrH TRIGA fuel. The method proves to be useful by quantifying the similarity with a large number of available benchmarks