Over the years, predictive modelling has been characterized as being environmentally deterministic, a-temporal, or even as a way of ‘effectively de-humanising the past’. Over the past ten years, however, spatial analysis of settlement patterns has progressed substantially, paying much more attention to the role of socio-cultural factors and the analysis of settlement pattern dynamics. In this paper, we will present an approach to site location analysis and predictive modelling that can be characterized as essentially data driven, yet is very much theoretically informed, and which has focused primarily on facilitating comparisons between various chrono-cultural contexts. Our experiments, that have been carried out since 2010, have mainly used data from the Roman period in various regions of France, but the general ideas and workflow can easily be transferred to other settings. To enrich the approach new developments were tested to understand the role of settlement hierarchy and its influence on the subsequent development and structuring of settlement patterns. These new developments were applied to three case study carried out in the north-east of France.