Data-driven real estate valuation, part II: data, uncertainty and governance for reliable automated valuation models in practice

Purpose This paper is part II of a two-part review of automated valuation models (AVMs). Part II examines what is required for AVMs to be reliable, fair and production-ready in high-stakes valuation practice, focussing on validation, uncertainty quantification, market coverage, data enrichment including environmental, social and governance (ESG) and synthetic data, fairness and bias, and governance. Part I covers the methodological foundations, the AVM pipeline and the main model families. Design/methodology/approach The paper follows the same narrative, practitioner informed review approach set out in Part I, synthesising peer-reviewed literature, binding regulatory and professional standards, and publicly available AVM provider documentation. Findings are consolidated in an evidence map that distinguishes established, emerging and speculative claims. Findings The review identifies four findings. First, data quality and granularity set the upper bound of AVM performance more fundamentally than model complexity. Second, calibrated prediction intervals and periodic backtesting are prerequisites for high-stakes use. Third, ESG signals and synthetic data expand the AVM data frontier but remain governance-sensitive. Fourth, hybrid human and AVM architectures prove most resilient, aligning with empirical and regulatory expectations. Practical implications Providers should attach calibrated prediction intervals to every AVM output. Users should monitor performance through periodic backtesting and bias audits. Regulators should treat AVM governance as part of broader model risk management. Originality/value The paper offers an integrative synthesis of data, uncertainty and governance in AVMs and positions hybrid human and AVM architectures as the most defensible operating model under current regulatory expectations.