The aim of this study is to estimate the bias of the adjusted gender wage gaps when productivity is not directly measured. In this study we rely on a unique data set with productivity and wages for 2,292 employees in a Polish fashion retail firm. We apply parametric decomposition methods to evaluate the gender wage gap with and without productivity measures. Our results suggest that the bias stemming from an omitted variable may indeed be significant, sometimes even altering the direction of the conclusions. Most of the estimates of the gender wage gap adjusted for individual characteristics do not account for productivity indicators due to a lack of such indicators in most available data sets. Although our results are obtained using data for only one employer, they suggest that most of the adjusted gender wage gap estimates in the literature may be substantially biased.