This paper aims to verify the presence of the learning‐by‐exporting effect on total factor productivity growth. The study starts, as is typical in this context, by addressing the pre‐entry selection bias at firm level but differs from the literature by focusing on the distribution of the outcome and considering the presence of the different influences of macroeconomic factors on exporters and non‐exporters. Additionally, the paper addresses the panel attrition, a current source of estimation bias in longitudinal studies. The analysis is based on a panel of Italian manufacturing firms in the 1998–2007 period. We design an experiment by aligning and pooling cohorts of firms that allow us to obtain a sufficiently large group of firms entering the international market. Our results show that internationalisation affects firms' productivity and that the effect is heterogeneous over total factor productivity distribution and larger for the firms at the bottom section of the distribution itself. Furthermore, we observe that the learning‐by‐exporting effect may be confounded without (a) considering that domestic and exporter firms may afford heterogeneous demand cycles and (b) managing the dropout of some firms from the panel.