Are present categorisations suited to the analysis of industrial change? To this question, we answer that while the light shed by national statistics is undeniable, the construction of these categorisations does not make it possible to segment economic activity homogenously or to deduce from it an industrial framework on which to base analysis of recent developments. Indeed, we first show that the historical bases for categorisations remain largely rooted in a historical reality, that of the industrial revolution, in which industry was particularly synonymous with technical progress and job creation. While profound technical and organisational changes have taken place over the last twenty and more years, the article underlines the importance of re-evaluating the industrial landscape, taking into account both its heterogeneous character and the need to analyse industrial activity with a constant perimeter. From this perspective, we highlight the main difficulties encountered by category-formers and economists. We then suggest three approaches which correspond to three definitions and three fundamental questions for the economist. Finally, we empirically illustrate our approach, applying these definitions to the 1978-2003 period, and suggest avenues to explore in order to undertake further research in this area.