No-Value Policy
An unreliable valuation is worse than no valuation. When data is insufficient to produce a meaningful estimate, Gadsden Meridian declines rather than guessing.
When we decline to value
The model returns a “no value” response when any of the following conditions are met:
- Insufficient comparable transactions. Fewer than 3 comparable sales within 2km in the trailing 5 years. The model requires a minimum density of local evidence to produce a meaningful estimate.
- Missing critical property data. No EPC record, unknown property type, or no floor area available. These are foundational inputs — without them, the model cannot reliably distinguish between properties.
- Unusual or unique property types. Properties where the model has no reliable training data — for example, large estates, commercial conversions, or highly bespoke homes.
- Confidence below minimum threshold. Even if data is technically available, the model’s confidence score may fall below the threshold required for a meaningful estimate.
Why we decline
An unreliable number creates false confidence
A valuation returned without adequate evidence can be worse than no valuation at all. Lenders and underwriters may rely on a figure that looks authoritative but isn’t supported by data, leading to mispriced risk in the loan book.
Regulators and rating agencies expect it
Fitch, Moody’s, and the PRA all require confidence measures precisely because not all properties can be reliably valued by AVM. A model that always returns a number is a model that doesn’t know what it doesn’t know. Confidence scoring and no-value policies are quality signals, not coverage failures.
Our no-value rate is a quality metric
We track no-value rates by geography, property type, and data availability. This data drives continuous improvement in data coverage and model accuracy. A low no-value rate earned through better data is more valuable than a low no-value rate achieved by lowering standards.
What happens when we return “no value”
Structured API response
The API returns a structured response with status: "no_value"
and a reason field explaining why. Your integration
can route the property to the next step in your cascade automatically.
No charge
No-value responses are not charged. You only pay for successful valuations where the model has sufficient confidence to produce a reliable estimate.
Route to next option
The lender can route the property to their next option in the valuation cascade: another AVM, a desktop review, or a physical survey. The reason code helps determine the appropriate next step.
Example no-value API response
{
"status": "no_value",
"valuation": null,
"reason": "insufficient_comparables",
"reason_description": "Fewer than 3 comparable sales within 2km in trailing 5 years",
"missing_data": [],
"property_id": 12345,
"model_version": "v6.11"
}Improving coverage
No-value responses are not permanent. There are several ways coverage improves over time:
- • Supply an EPC. If a property falls short due to missing data, the client can supply an Energy Performance Certificate. EPCs provide floor area, construction age, and habitable rooms — often sufficient to move the property from “no value” to a confident estimate.
- • Growing transaction database. As new Land Registry, EPC, and listing data is ingested monthly, coverage improves continuously. Areas that lack sufficient comparables today may become coverable with the next data update.
- • Model improvements. Each model iteration improves feature engineering and data enrichment, reducing the number of properties that fall below the confidence threshold.
No-value reason codes
Each no-value response includes a machine-readable reason code for automated routing:
| Code | Description | Suggested action |
|---|---|---|
| insufficient_comparables | Fewer than 3 comparable sales nearby | Desktop review or physical survey |
| missing_critical_data | Key property attributes unavailable | Supply EPC or property details |
| unusual_property_type | Property outside model’s training data | Physical survey recommended |
| below_confidence_threshold | Model confidence too low for reliable estimate | Alternative AVM or desktop review |
Test our coverage on your portfolio
Upload your portfolio via backtest to see our coverage rate and accuracy on your specific properties. Free, no commitment.