Downloads
Published accuracy reports, technical documentation, and commercial materials. Everything your risk committee needs.
Quarterly Accuracy Report
Full accuracy metrics from our latest bulk test of 295,026 Land Registry transactions. Headline metrics, segmented results by property type, price band, and region, plus known limitations.
Meridian v6.12.1 • H2-2025
Technical Summary
Model architecture, data sources, feature categories, validation methodology, confidence model, regulatory position, and known limitations. For risk managers and model risk teams.
Meridian v6.12.1
One-Pager
Single-page executive summary with key metrics, pricing, and contact details. Forward to colleagues, decision-makers, or procurement teams.
Model Risk / Vendor Due Diligence
The Gadsden Valuations vendor due-diligence pack: three issued PDFs — dated, version-stamped, and regenerated only when a model or testing change occurs.
Currently issued: pack v2.1, 14 June 2026, against Gadsden Meridian v6.12.
Gadsden Meridian — Technical Documentation
Full model methodology, data sources, validation, accuracy results, the FSD-to-coverage mapping required by EAA ESSVM §5.3, explainability and audit trail, the v6.13 development cycle, and the API annex. The detailed document a model-risk team works through.
Accuracy Evidence Summary
Standalone tear-out of the H2-2025 census results — headline metrics, the A/B/C/D band table, by-type/price-band/region segmentation. The pack's only volatile-numbers document; reissued with every bulk test.
Company Facts
Legal entity, regulatory status, terms, service and security — the page procurement asks for.
Pack v2.1 carries v6.12 (recalibration release of 14 June 2026 on the v6.11 prediction model). Older issues are superseded by this one and are not republished.
Sample Reports
Real valuations of real UK properties, marked SAMPLE on every page. The set is deliberately contrasting — strong evidence, weak evidence, and a property the model declined to value — so you can see exactly how the report behaves before you buy.
Fitch Band A
No sample available — the model has not yet produced a Band A valuation (none in the current test). When it does, a sample will appear here.
Fitch Band B — Strong Evidence
Dense local sale and rental evidence produces a tight valuation range and a Band B confidence rating. BS3 1FH, valued at £319,416.
Ref GV-20260611-00007
Fitch Band C — Standard Case
A typical mid-band result: solid sale comparables with thinner rental evidence, mapped to Band C. B13 0DW, valued at £276,065.
Ref GV-20260611-00008
Fitch Band D — Honest Weak Case
Conflicting evidence produces a wide range (£52,571 – £178,004), disclosed rather than hidden — Band D. CF24 0DH.
Ref GV-20260611-00009
No-Value Response
What you receive when the model declines to value rather than return an unreliable figure: the decline reason and guidance to route the property onward in your valuation cascade. No valuation is consumed.
Ref GV-20260611-00010
What's in the report
Property Details
The report begins with the property's address, postcode, type, floor area, and EPC rating (where available). This data is sourced from the national EPC register and Land Registry records.
Estimated Market Value
The headline figure: a point estimate of current market value in pounds sterling. Below it, a prediction range shows the upper and lower bounds reflecting model uncertainty for this specific property.
Confidence Assessment
Every valuation carries a per-property Forecast Standard Deviation (FSD) based on comparable evidence quality, data completeness, and prediction certainty, mapped to a Fitch A–D classification band. A lower FSD means stronger evidence and a tighter range.
Comparable Properties
A table of nearby recent sales used by the model, showing address, distance from the subject property, sale date, sale price, property type, and floor area. This gives you the same evidence a surveyor would use for a comparison-based valuation.
Feature Contributions
An explainability breakdown showing which property characteristics pushed the valuation up or down relative to the local median. This is derived from SHAP (SHapley Additive exPlanations) analysis of the model's decision.
Methodology & Disclaimer
Every report includes a plain-language explanation of the model methodology, data sources used, and a clear disclaimer stating this is an AVM estimate, not a formal valuation. Full data provenance information for regulatory and due diligence purposes.