This paper proposes a semi-nonparametric (SNP) generalization of the lognormal distribution for studying firm size and providing accurate measures for economic concentration and inequality in terms of the Gini index adjusted to flexible functional forms. The empirical application for a sample of 1,772 Colombian companies from 2002 to 2015 shows that the log-SNP provides a better fit to firm size distribution, especially for the extreme quantiles, for which lognormal distribution overestimates economic concentration. In addition, dynamic panel model estimates indicate that firm characteristics, including size, age, and leverage, are determining factors in explaining firm growth, thus rejecting Gibrat’s law.