With reactivity being the most important integral reactor physics quantity - and simultaneously the one that can be measured with the highest accuracy - there is a great interest in understanding how possible space- and energy-dependent data and/or modeling discrepancies may propagate into a calculated reactivity change, and with which magnitude this occurs. In the context of pin removal reactivity effects in a light water reactor assembly, for example, it is illustrative to carry out, for any arbitrary localized material composition perturbation, a decomposition of the total effect into individual space- and energy-dependent contributions of the different unit cells in the assembly. If this decomposition is normalized to +100% in the case of a positive reactivity effect and to -100% in the case of a negative reactivity effect, an importance map is established that indicates the relative contribution (in percent) of each individual contributing cell to the total reactivity effect caused by the localized material composition change. Such an importance map can be interpreted as a sensitivity matrix that quantifies the final discrepancy in a calculated reactivity effect, with respect to its reference value, as a weighted sum of the complete collection of cell-wise data and/or modeling discrepancies. The current paper outlines the basic theory and gives certain practical applications of the proposed decomposition methodology. Thus, it is found that the developed methodology offers in-depth, quantitative explanations for calculational discrepancies observed in the analysis of fuel pin removal experiments conducted in the framework of the LWR-PROTEUS program at the Paul Scherrer Institute.