Built for Regulated Lending
Designed for underwriting departments and model risk teams. PRA aligned, independently verifiable, fully auditable. Can you put this in front of your risk committee?
Gadsden Valuations is a PRA SS1/23-aligned automated valuation model with published accuracy metrics and walk-forward validation against real Land Registry sale prices. Per-property confidence scoring, full audit trail, and transparent methodology — built for lender underwriting requirements and model risk team due diligence.
PRA SS1/23
The Prudential Regulation Authority’s Supervisory Statement SS1/23 sets expectations for how regulated firms should use model-driven valuations in lending decisions. While Gadsden Valuations is not a PRA-regulated firm, our model and reporting are designed to support lenders’ compliance with SS1/23 requirements, including:
- • Model transparency: Published methodology, training data, and feature descriptions.
- • Accuracy monitoring: Regular, published backtest results against realised sale prices.
- • Confidence indicators: Per-valuation confidence tier enabling risk-based decision-making.
- • Audit trail: Each valuation is recorded with model version, input data, and timestamp.
Basel 3.1 readiness
Basel 3.1 implementation (effective 1 January 2027 in the UK) introduces stricter requirements for property revaluation in mortgage risk-weight calculations. Firms will need scalable, independently validated property valuations with documented accuracy. Our model supports this transition:
- • Portfolio-scale revaluation: Batch CSV upload or API for thousands of properties per hour.
- • Per-property confidence: Each valuation is individually scored, supporting risk-proportionate treatment.
- • Published backtest metrics: PE10, MdAPE, and segmented accuracy data available for model risk assessment.
- • Audit trail: Every valuation is versioned and reproducible for regulatory examination.
Ratings agency methodology alignment
For RMBS and securitisation use, rating agencies (Fitch, Moody’s, S&P) review AVM vendors against their own criteria. The metrics they evaluate — PE10, Forecast Standard Deviation, and per-property confidence — are the same metrics we publish openly.
- • PE10 (PPE10): The primary accuracy metric used by Fitch for AVM vendor assessment. Published on our accuracy page.
- • Forecast Standard Deviation: Per-property confidence measure used by rating agencies to determine haircut bands for RMBS transactions.
- • Transparent methodology: Full model documentation available for vendor due diligence, structured to support the review process.
- • Benchmarked against sale prices: We test against actual Land Registry completion prices, not surveyor estimates — a harder and more honest benchmark.
EAA ESSVM alignment
The European AVM Alliance (EAA) publishes the European Standards for Statistical Valuation Methods (ESSVM), which provide a framework for AVM quality, governance, and transparency. Our approach aligns with the ESSVM requirements in the following areas:
- • Model governance: Documented model development, versioning, and change management.
- • Performance monitoring: Published accuracy metrics with regular backtesting against new sale data.
- • Confidence scoring: Three-tier confidence classification with clear definitions.
- • Coverage reporting: Clear disclosure of geographic and property-type coverage.
Professional standards context
Our methodology is consistent with the principles set out in relevant professional valuation standards. These provide useful context for how automated valuations fit within the broader property valuation framework.
IVS 105 — Valuation Approaches and Methods
The International Valuation Standards (IVS) 105 sets out the three recognised valuation approaches: market, income, and cost. Statistical models fall within the market approach when they are calibrated to actual transaction evidence. Our model is trained exclusively on realised sale prices from HM Land Registry, placing it firmly within the market approach as defined by IVS 105.
RICS PS 1.3 — Statistical valuation
RICS Professional Standard 1.3 recognises statistical valuation methods as a legitimate approach to estimating market value. It requires that statistical models disclose their methodology, accuracy, data sources, and limitations. Our valuation reports and methodology documentation are designed to meet these disclosure requirements.
RICS PS 1.6 — Automated valuation models
PS 1.6 addresses the use of AVMs specifically. It requires that AVM providers publish accuracy metrics, disclose confidence levels, identify the model type, and maintain an audit trail. Our confidence tier system, published backtest metrics, model version tracking, and per-valuation audit records are consistent with these requirements.
Data provenance
All input data is sourced from government and public-sector bodies under their respective open data licences. We do not use scraped portal data, asking prices, or self-reported property details.
| Source | Provider | Licence | Update frequency |
|---|---|---|---|
| Price Paid Data | HM Land Registry | OGL v3.0 | Monthly |
| EPC Register | MHCLG | OGL v3.0 | Monthly |
| House Price Index | ONS / Land Registry | OGL v3.0 | Monthly |
| School data | DfE / Ofsted | OGL v3.0 | Termly |
| Crime statistics | Police.uk / Home Office | OGL v3.0 | Monthly |
| Flood risk | Environment Agency | OGL v3.0 | Quarterly |
GDPR and privacy
We take data protection seriously. Our processing activities are conducted in accordance with the UK General Data Protection Regulation (UK GDPR) and the Data Protection Act 2018.
Personal data we process
Account registration details (name, email). Valuation requests (property addresses). Payment details are processed by Stripe and never stored on our servers.
Lawful basis
Contract performance (providing the valuation service you requested) and legitimate interests (improving model accuracy, fraud prevention).
Data retention
Valuation records are retained for audit purposes. Account data is retained while your account is active and for 6 years thereafter (HMRC requirement). You may request deletion of your account at any time.
Your rights
You have the right to access, rectify, erase, restrict processing, and port your personal data. Contact us at privacy@gadsdenvaluations.co.uk.
Disclaimer
Gadsden Valuations is an automated valuation model (AVM) that estimates UK residential property market value using a machine learning model trained on publicly available government data. The model is validated through walk-forward backtesting against real Land Registry sale prices, with accuracy metrics published transparently on our accuracy page. Each valuation includes a confidence tier and prediction interval to support risk-based decision-making.
Gadsden Valuations estimates are not RICS Red Book valuations and are not produced by a RICS-registered valuer. They should not be used as a substitute for a formal valuation where one is required by a lender, insurer, court, tax authority, or other regulatory body, nor relied upon as the sole basis for property purchase, sale, lending, or investment decisions.
All model-based estimates carry inherent uncertainty. The confidence tier and prediction interval provided with each valuation should be considered alongside the point estimate.
Gadsden Valuations Ltd accepts no liability for losses arising from reliance on AVM estimates. Users are advised to obtain independent professional advice where appropriate.