Start Date
26-6-2025 9:15 AM
End Date
26-6-2025 10:30 AM
Description
The Valuation Office Agency (VOA) values 2.1 million non-domestic properties for Business Rates every three years. Each Revaluation faces pressure to be more efficient than the last, however the VOA must always deliver a high-quality Rating List that is defensible in tribunals and court. In acknowledgement of these competing priorities, our valuation modelling team has focused our research on those statistical models that provide not just first pass valuations, but explainable ones. We hope that this will increase the VOA’s ability to deliver a Revaluation that is more efficient, consistent, and trusted. This presentation will share our work looking into a promising statistical model structure for shop valuations – Whittle-Matérn Fields on Metric Graphs – which aligns well with traditional valuation methods whilst reaping the benefits of an automated data driven approach. We hope that this session will be of interest to others grappling with the problem that valuation is sometimes just as much about the justification as the final answer.
Recommended Citation
Razzelliot, Matt and Vaughan-Williams, Alice, "Prioritising explainability in non-domestic AVM research" (2025). Mass Appraisal Valuation Symposium. 15.
https://researchexchange.iaao.org/mavs/mavs2025/sessions/15
Prioritising explainability in non-domestic AVM research
The Valuation Office Agency (VOA) values 2.1 million non-domestic properties for Business Rates every three years. Each Revaluation faces pressure to be more efficient than the last, however the VOA must always deliver a high-quality Rating List that is defensible in tribunals and court. In acknowledgement of these competing priorities, our valuation modelling team has focused our research on those statistical models that provide not just first pass valuations, but explainable ones. We hope that this will increase the VOA’s ability to deliver a Revaluation that is more efficient, consistent, and trusted. This presentation will share our work looking into a promising statistical model structure for shop valuations – Whittle-Matérn Fields on Metric Graphs – which aligns well with traditional valuation methods whilst reaping the benefits of an automated data driven approach. We hope that this session will be of interest to others grappling with the problem that valuation is sometimes just as much about the justification as the final answer.