This article scrutinizes the administrative and scientific practices by which the Gypsy/Roma category has been historically constructed in Europe, particularly in Central and South Eastern Europe. Censuses, police-led inquiries, social surveys and expert estimates made the ‘Roma’ category appear objective, making invisible the multiple technical and political decisions behind their interpretations. This paper examines the technologies behind the production of ethnic categories and that of Roma in particular: namely, guidelines and manuals for field workers, recruitment and census campaigns, consent forms, questionnaires, data processing, basic assumptions and interpretation of data. While expert networks give objectivity and flexibility to the collection and circulation of data, the labour involved in crafting ethnic statistics often remains obscure. This paper follows the historical departures and continuities of Roma categorization from the 18th century to present times. The category ‘Roma’ was produced and reproduced through quantification practices with the justification that ethnic data (or categorization) would help solve social problems and contribute to Roma integration. The technical literature reviewed in this paper and the auto-ethnographic analysis shed light on the machinery of ethnic categorization, and allows us to assess the impact of various kinds of labour upon this categorization: from the more visible work performed by ethnopolitical entrepreneurs and scholars – to the relatively invisible contributions of field workers, administrators and census takers. This article calls for a critical scrutiny of how Roma ethnicity is crafted through practices of ethnic quantification and encourages researchers to use methodological prudence and more self-reflection in their own academic practices.