The emancipatory possibilities brought by computational methods to digital archives are enormously exciting, not least in the building of historical population databases. Around the world, and especially in the Nordic countries, demand for historical population databases is driven by researchers in diverse fields including history, the social sciences, and medicine. Designed to provide individual and multigenerational data, these databases extend registry-based research back several generations before the introduction of modern registration systems. However, the term “data” is often positioned in relation to neutrality and objectivity and frequently discussed in the language of the natural sciences as a raw material. In transforming digitised collections into data, we address potential oversights when exploiting new computational opportunities. This chapter takes the Denmark-based Link-Lives project, housed at the Danish National Archives, as our empirical point of departure to analyse the multiple steps in source selection and transformation of non-digital records into research data, addressing the different logic, motivations, and challenges (human and algorithmic) as well as the actors involved and the consequences of each. We discuss this using FAIR principles and propose a model that captures the various complexities. Our work is of interest especially to those engaged in transforming archival records into research datasets.