Reference

How Zyfy Intelligence Works

The raw government data behind every signal is public. This is the layer Zyfy builds on top of it — a real methodology, not a black box.

1. Raw ingestion

Every signal starts from an authoritative UK source — DVLA, DVSA, ONS, Environment Agency, HMLR, Ofcom, Police UK, Ordnance Survey, BGS, DEFRA, UK Parliament, Euro NCAP — pulled directly under the Open Government Licence v3.0. Not a data broker resale, not a third-party aggregator sitting between Zyfy and the source.

2. Normalisation

Source formats vary — bulk CSV, GeoPackage, REST APIs each with their own quirks. Each gets reduced to a consistent postcode- or vehicle-keyed signal, with a sensible fallback when the underlying source data is sparse rather than simply returning null. See Coverage & Limitations for where those fallbacks apply.

3. National percentile ranking

Every scoreable signal is ranked against the full national distribution, not just reported as a raw government figure. A flood risk of "medium" tells you a category; knowing where that sits relative to the rest of Great Britain tells you something more useful. Raw percentiles are exposed directly via the percentiles object for anyone building their own model on top.

4. Composite scoring

Scores like propertyRiskScore and liveabilityScore combine multiple national percentiles into a single weighted result — flood risk and crime feed property risk; air quality, deprivation, and green space feed liveability; and so on for each score. The weighting reflects which factors matter most for that particular judgment, tuned against real outcomes rather than picked arbitrarily.

5. Graceful degradation

When a component signal is unavailable, Zyfy redistributes rather than nulling the whole score outright — a single missing input doesn't take down an entire composite score. Scores only return null when there genuinely isn't enough underlying data to say anything meaningful.

6. Confidence signalling

summary.dataConfidence and signals.property.trendConfidence tell you how much weight to put on a given result, based on how much underlying data was actually available. A result flagged with lower confidence should be treated with more caution in anything built downstream.

7. Not predictions

Every composite score describes the present state of a postcode or vehicle against the national distribution — it does not forecast the future. summary.investmentOutlook, summary.growthSignal, and every other score in the scores.* and summary.* families are derived, structured signals for you to weigh — not a guarantee of what happens next.