This contribution conducts a critical review of the many features of the traditional regression approach to domestic violence that implies a target variable (intimate partner femicide) and multiple predictors. Based on conceptual, theoretical, methodological and statistical considerations, this analysis, first, addresses the issue of how the risk factors can be properly organized; second, it claims that what are currently regarded as the independent variables might have numerous different explanatory roles, making reference to direct, indirect, mediated or moderated models, as well as to the facilitating and suppressing effect of variables; finally, the article introduces the concept of sequential behavioural patterns. Assuming that the build-up to partner femicide (PF) is a dynamic process, the analysis cautions against ignoring the sequence through which the different moves among the actors actually take place. The same events occurring in a different order may result in entirely different outcomes. This contribution is essentially theoretical, implementing worked examples from literature on risk factors and femicide. Implications for researchers and operators in the field of PF are underlined.