Data-driven real estate valuation, Part I: foundations, pipeline and model families of automated valuation models

Purpose This paper is Part I of a two-part review of Automated Valuation Models (AVMs) as data-driven tools for scalable property valuation. Part I focusses on the methodological foundations of AVMs, the AVM pipeline and the main model families and situates them within the international regulatory landscape and the transparency versus performance trade-off. Part II addresses data foundations, uncertainty, fairness and governance. Design/methodology/approach The paper is a narrative, practitioner-informed review that synthesises peer-reviewed literature, binding regulatory and professional standards across the European Union (EU), the UK and the USA and publicly available AVM provider documentation. AVMs are classified along three dimensions, namely data basis, modelling approaches and application logic and their properties are compared across model families. Findings Three findings run through the review. First, regulatory frameworks remain fragmented across jurisdictions even as AVMs increasingly influence valuation decisions despite their formal status as supporting tools. Second, machine learning has expanded methodological capabilities, while interpretability increasingly determines regulatory acceptance. Third, semi-parametric approaches such as Generalised Additive Models and Generalised Additive Models for Location, Scale and Shape emerge as a methodological sweet spot between hedonic transparency and machine-learning flexibility. Practical implications Model designers and validators can use the comparative overview in Table 1 to align model choice with the interpretability, data and regulatory requirements of a given valuation task. Originality/value The paper offers an integrative taxonomy of AVM model families that links methodological properties to the recent regulatory turn in artificial intelligence and AVM oversight.