RICS & UK Regulation

RICS Red Book Global Standards

January 2025 edition — effective 31 January 2025

The RICS Valuation – Global Standards (commonly known as the Red Book) is the definitive framework governing professional valuation practice worldwide. The January 2025 edition is the first to address automated valuation models and artificial intelligence explicitly within the Professional Standards. It does not prohibit AVM use — it regulates it.

PS 1: AVM output plus professional judgement

Professional Standard 1 establishes the critical distinction: an AVM output on its own is not a Red Book compliant valuation. However, an AVM output combined with the professional judgement of a RICS Registered Valuer is a written valuation within the meaning of the standards. The valuer must review the model output, apply their expertise to account for factors the model cannot observe (internal condition, legal encumbrances, micro-location), and take responsibility for the final figure.

This framework positions AVMs as professional tools that augment valuer judgement rather than replace it. It creates a clear, standards-compliant pathway for AVM-assisted valuations at scale — provided the AVM output is robust, transparent, and accompanied by appropriate quality indicators.

PS 1.4: AI and open-source tools are not prohibited

PS 1.4 directly addresses the use of artificial intelligence and open-source tools in valuation practice. The standard is clear: the use of AI-powered models, open-source software, and commercial machine learning tools is not prohibited, provided they are subject to appropriate governance. This means documented model methodology, validated accuracy metrics, version control, and a clear audit trail. The standard recognises that modern valuation increasingly relies on computational tools and seeks to ensure quality governance rather than restrict technological choice.

PS 1.6: Specialist or service organisation

PS 1.6 permits a RICS Registered Valuer to engage a specialist or service organisation to provide AVM outputs as part of the valuation process. The valuer retains professional responsibility for the final valuation, but the model itself may be developed, maintained, and operated by a third-party provider with relevant technical expertise. This is the standard that permits the existence of commercial AVM providers serving the UK valuation market — the valuer does not need to build the model themselves.

VPS 5: New Valuation Models section

Valuation Practice Statement 5 in the January 2025 edition introduces a dedicated section on valuation models, implementing the requirements of IVS 105 within the RICS framework. VPS 5 requires that where a valuation is supported by a model, the valuer must understand the model’s methodology, be satisfied that it is appropriate for the subject property, and disclose the model’s role in the valuation process. This creates a direct link between the RICS Professional Standards and the IVS model quality requirements described below.

IVS 105: Valuation Models

International Valuation Standard 105 is the global standard governing the use of models in valuation practice. Its preamble establishes the foundational principle: no model, without a valuer applying professional judgement, can produce a valuation that complies with IVS. This is not an anti-technology statement — it is a quality assurance requirement that ensures human expertise remains central to the valuation process.

Sections 30–50: Model characteristics

IVS 105, sections 30 through 50, define four characteristics that valuation models must demonstrate. These requirements apply to any model used in the valuation process, including AVMs:

Accuracy

The model must produce outputs that can be tested against actual market evidence. Published accuracy metrics (PE10, MdAPE, prediction intervals) measured against realised transaction prices are the standard way to demonstrate this.

Completeness

The model must incorporate all material factors that influence value, or explicitly acknowledge those it cannot observe. Known limitations must be disclosed so that the reviewing valuer can adjust for unmodelled factors.

Timeliness

Input data must be current. The model must account for market movement between the date of the most recent comparable evidence and the valuation date. Temporal adjustment mechanisms (such as house price index adjustments) satisfy this requirement.

Transparency

The model’s methodology, data sources, and limitations must be documented and available to the valuer and to the client. Feature contributions, comparable evidence, and confidence indicators support this requirement.

IVS 105.20: The specialist provision

Section 20 of IVS 105 permits the valuer to engage a specialist or service organisation to provide the valuation model. The model provider must document their methodology, data sources, and limitations. The model must be tested for functionality, and its outputs analysed for accuracy. The valuer retains responsibility for ensuring the model is fit for purpose and for applying professional judgement to the output.

This provision is the international-level authority for commercial AVM services. It creates a clear division of responsibility: the AVM provider is responsible for model quality, documentation, and accuracy; the valuer is responsible for selecting an appropriate model, reviewing its output, and forming a professional opinion.

PRA SS1/23: Model Risk Management

Supervisory Statement SS1/23, published by the Prudential Regulation Authority in May 2023, sets expectations for how PRA-regulated firms should manage model risk. It applies to all models used in regulated decision-making, including automated valuation models used in mortgage lending and portfolio management.

SS1/23 does not regulate AVM providers directly — it regulates the firms that use them. However, its requirements have significant implications for what a regulated firm needs from its AVM supplier:

•
Model inventory and documentation: Regulated firms must maintain an inventory of all models in use, with documentation of each model’s methodology, assumptions, data inputs, and known limitations. AVM providers must supply documentation sufficient for the firm to complete its model inventory entry.
•
Independent model validation: Models must be subject to independent validation before deployment and on an ongoing basis. For third-party models, this means the regulated firm must be able to assess model performance against its own portfolio data, which requires the AVM provider to publish accuracy metrics and support backtesting.
•
Model governance: Clear ownership, version control, change management, and escalation procedures. Firms need to understand when the model changes and what the impact of those changes is on performance.
•
Third-party model risk: SS1/23 explicitly addresses third-party models. Firms cannot outsource model risk by outsourcing the model. They must conduct appropriate due diligence on external model providers and maintain ongoing oversight of model performance.

In practice, this means an AVM provider serving regulated firms must offer transparent methodology documentation, published and independently verifiable accuracy metrics, segmented performance data, and the ability for clients to test the model against their own portfolios.

FCA MCOB: Mortgage Valuation Requirements

The Financial Conduct Authority’s Mortgages and Home Finance: Conduct of Business sourcebook (MCOB) governs how mortgage lenders must approach property valuation. MCOB 5.6 requires lenders to undertake a “proper valuation” before entering into a regulated mortgage contract. The interpretation of what constitutes a proper valuation varies by transaction type.

Where physical valuation is typically mandated

Purchase transactions where the borrower is acquiring a new property generally require a physical valuation by a RICS Registered Valuer. Higher loan-to-value lending (above 75–80% LTV depending on lender policy) also typically triggers a physical inspection requirement. First-time buyers, non-standard construction, and properties in areas with limited comparable evidence are further cases where physical inspection is standard practice.

Where AVM-assisted valuation is permitted

Remortgage transactions, particularly at lower LTV ratios, are the primary use case where AVM-assisted valuations are accepted by many lenders. Product transfers (where the borrower remains with the same lender) frequently use AVM valuations. Portfolio revaluation for capital adequacy purposes, further advances on existing mortgages, and equity release initial screening are further accepted use cases for AVM-based approaches.

The regulatory position is evolving. Lenders have significant discretion in determining when an AVM is appropriate for their risk appetite, and several major UK lenders now accept AVM-assisted valuations for a growing proportion of their remortgage book. The trend is towards wider AVM acceptance, driven by cost efficiency, speed, and the improving accuracy of modern models.

Basel 3.1: The Coming Demand Catalyst

Expected implementation: January 2027 (UK)

The Basel 3.1 framework (CRR3 in EU terminology, implemented in the UK via PRA rules) introduces requirements that will significantly increase demand for automated property valuations across the UK banking sector. Two provisions are particularly relevant:

Portfolio revaluation requirements

Basel 3.1 requires banks to revalue their property collateral on a regular basis for capital adequacy calculations. For residential mortgages, this means the entire mortgage book must have current valuations, not just the original purchase or remortgage valuation. Physical revaluation of millions of properties is neither practical nor proportionate. AVMs are the only viable solution for portfolio-scale revaluation, and the PRA has indicated that properly governed AVM use is acceptable for this purpose.

The 10% threshold monitoring requirement

Under Basel 3.1, banks must monitor their property collateral for value movements and trigger a revaluation when there is reason to believe the property value has changed materially — commonly interpreted as a 10% decline from the last recorded value. This creates a requirement for continuous or frequent automated monitoring of property values across entire portfolios, not just ad hoc revaluation.

The scale of the requirement is substantial. UK lenders collectively hold mortgage books covering millions of residential properties. Each of those properties will need regular automated valuations to satisfy Basel 3.1 capital adequacy rules. This creates a structural, regulation-driven demand for AVM services that is independent of the existing mortgage origination market.

Firms that have not yet procured AVM capability for portfolio revaluation will need to do so before the January 2027 implementation date. The PRA expects firms to have arrangements in place well in advance.

The Explainability Requirement

A persistent misconception in the AVM market is that regulators require models to be fully interpretable — that “black box” models are prohibited. This is not what the standards say.

The European AVM Alliance’s ESSVM language, and the broader regulatory stance reflected in IVS 105 and PRA SS1/23, assumes that black-box elements exist within sophisticated valuation models. Gradient-boosted trees, neural networks, and other machine learning approaches inherently contain complex non-linear interactions that cannot be reduced to simple formulae. The standards accept this.

What the standards require is explainability alongside the model, not instead of it. The requirement is that the model’s outputs can be explained to a professional user — that the valuer can understand why the model produced a particular estimate for a particular property. This is achieved through:

• Feature contributions: showing which input variables had the greatest influence on the output (SHAP values, feature importance rankings)
• Comparable evidence: presenting the actual transactions the model considers most similar to the subject property
• Confidence indicators: communicating the model’s own assessment of estimate reliability
• Prediction intervals: providing upper and lower bounds that reflect the range of likely values

The practical implication is that a well-governed machine learning model with robust explainability outputs is fully compatible with current regulatory requirements. The requirement is for transparency of output, not simplicity of method.

How Gadsden Valuations Aligns

Gadsden Valuations is designed from the ground up to satisfy the requirements described on this page. Every architectural decision — from published accuracy metrics to per-valuation feature contributions — exists because a specific standard requires it.

Published accuracy metrics

PE10, MdAPE, and prediction intervals measured against Land Registry sale prices via walk-forward backtesting. Segmented by property type, price band, and region. Satisfies IVS 105 accuracy requirements and supports PRA SS1/23 model validation. View our accuracy metrics.

Confidence tiers per IVS 105

Three-tier confidence classification (High, Medium, Low) with an explicit Declined category for properties with insufficient data. Mapped to the EAA common 0–7 Forecast Standard Deviation scale. Each valuation communicates its reliability level so that the reviewing valuer or lending policy can respond appropriately.

Explainability: comparable evidence and feature contributions

Every valuation includes the comparable transactions the model considers most relevant and a breakdown of which property features contributed most to the estimate. This satisfies the IVS 105 transparency requirement and the ESSVM explainability expectation without requiring the model itself to be a simple linear formula.

Walk-forward backtesting methodology

All published accuracy figures use strict walk-forward validation: the model is trained on historical data and tested exclusively on future transactions it has never seen. This prevents data leakage and ensures metrics reflect genuine predictive performance. The methodology is documented and the test set is defined by date cutoff, not random sampling.

Documentation for regulated firm due diligence

Technical methodology documentation, model governance procedures, and data source provenance are available to support the vendor assessment process required by PRA SS1/23. Read our Technical Summary.

Regulatory Standards at a Glance

Standard	Applies to	Key AVM requirement
RICS Red Book (Jan 2025)	RICS Registered Valuers	AVM output + professional judgement = compliant valuation
IVS 105	Valuation model providers	Accuracy, completeness, timeliness, transparency
PRA SS1/23	PRA-regulated firms	Model governance, validation, documentation for third-party models
FCA MCOB	Mortgage lenders	Proper valuation required; AVM permitted for qualifying transactions
Basel 3.1 (Jan 2027)	Banks with mortgage portfolios	Portfolio revaluation and 10% threshold monitoring

Built for Regulatory Scrutiny

Gadsden Valuations publishes its methodology, accuracy metrics, and confidence model because the standards require it. Read the full technical summary to understand how the model works.

Read our Technical Summary