A simplified composition-independent atomic size difference factor is proposed to account for the statistical deviation of original Miedema's model for formation enthalpies of binary intermetallic compounds from experimental data. We demonstrate that with the incorporation of this new factor, one can achieve a high level of correlation with respect to available experimental data. Using this statistically refined model, we propose a methodology for assessing where the next level of statistical refinement is needed, utilizing a data-mapping framework of 'critical descriptors'. This provides a way to identify the search space where needs further experimental work to be performed. (C) 2014 Elsevier B.V. All rights reserved.