Methodology 1.3.0
Every value comes from a named public source, is placed in national context, carries its own confidence, and is stamped with the version that produced it. So you can trace it, cite it, and get the same answer when you check it again.
01 · The primitive
A signal is one measured attribute of a UK area, from a named source, placed in national context, time-stamped and confidence-rated. Scores, comparisons and forecasts are all built on top of signals.
The measurement itself, in its own unit, or empty with a reason when there is nothing solid to report.
Where the value sits against comparable areas, from 0 to 100, so a number means something on its own.
How solid the number is, based on the source, the sample and how recent it is. Honest, not aspirational.
The named public dataset it came from, so you can cite it and stand behind it.
The engine version that produced it, so the same request returns the same number, months later.
One signal, with every part on show. Nothing is a bare number.
02 · Where the data comes from
Seven categories, each from official, public datasets. We name every one here, and every response carries its source alongside the value, so you can always show your working.
police.uk and the Home Office. England · Wales · Scotland.
IMD 2025, WIMD and SIMD. England · Wales · Scotland.
ONS HPSSA and HM Land Registry. England · Wales.
Ofsted. England.
OpenStreetMap. Great Britain.
NaPTAN. Great Britain.
Environment Agency and SEPA. England · Scotland.
Postcodes are resolved to neighbourhoods through the official ONS spine, so every value lands on the right area.
03 · National context
Every value is ranked against comparable areas. Each country is ranked within itself, because their deprivation indices are built differently and are not comparable across the border. You can also rank an area against its own region, so a strong area in a quieter part of the country is not flattened by the national picture.
National ranking · median price, England
Kensington and Chelsea
Westminster
Westminster
Westminster
Kensington and Chelsea
Regional ranking · top of each region
Kensington and Chelsea · London
Oxford · South East
Broadland · East of England
Oadby and Wigston · East Midlands
Northumberland · North East
The same signal, the same country, the same shortlist. National returns five slices of prime London; regional returns the top of five different regions.
04 · Historical snapshots
Each month we add a fresh snapshot of every signal in every area. We only ever add to that history, never overwrite it, so past values stay exactly as they were measured and any figure can be reproduced months later.
A correction shows up as the next month's value, so a number you saw before stays exactly as it was, and stays reproducible.
Because the history never changes, a figure you cite today returns the same figure when you or an auditor check it months from now.
The record gets deeper every month, building up area history that nobody can backfill after the fact.
05 · Scoring
A score turns an area's signals into a single number from 0 to 100. Every profile scores the same seven categories; the profile only changes how they are weighted for its job. The scoring is deterministic, so the same inputs always return the same score.
How liveable is this area for a household. Drives the area-quality lens on listing detail pages, valuation flows and relocation tools.
Where to open. Weights the categories toward commercial demand: spending power, commercial costs, amenities, transport. Drives shortlisting at portfolio scale.
What this area looks like as an asset. Growth trajectory, yield, regeneration context, tenant demand, downside risk.
Analyst-friendly default. Balanced weights across all seven categories. Survives FOI and procurement review.
You can re-weight the seven categories for a single request, or save a weighting against your organisation and reuse it. Every score is stamped with the version that produced it, so you can pin it and reproduce the exact number later.
06 · Beyond a single reading
As well as today's numbers, we work out how areas are changing, how they compare with their peers, and where a signal is heading. Each one comes with its own time window and its own confidence.
How prices, sales activity and crime have moved against a year ago.
The recent direction of travel, so a fast-moving area shows up early.
The longer-run direction over a couple of years, steadier than a single jump.
How an area reads against similar areas, high or low for its kind.
The areas most like a given one, based on how closely they sit across the signals they share. A simple, symmetric similarity.
Areas that stand out from their peer group on a signal, so unusual places surface on their own. Worked out ahead of time, so it is fast.
A straightforward projection of where a monthly signal is heading, with a confidence band around it. A transparent trend, not a black box.
07 · Ask a question
Ask a question in plain English and it is turned into a precise query. You get the answer and the exact query behind it, so every result can be reviewed and run again. Prefer to be exact? Send the typed query yourself and skip the AI entirely.
Your question, in plain English
English neighbourhoods under £250k where prices are rising, crime is below average and deprivation is low, best growth first, top five.
Understood as
Rank English areas, keep the ones under £250k with prices rising, crime and deprivation both in the better half, sort by growth, return the top five, and hand back the plan so you can run it again.
08 · Confidence
Every value comes with a confidence level and a short, plain-English reason. It reflects how fresh the data is, how large the sample is, whether we had to fall back to another source, and how much the signal moves around. When the data is thin, the response says so.
Fresh data from the primary source, a healthy sample, little movement. For example: Recent crime, or prices where there are plenty of sales.
An older release, a smaller sample, a fallback source, or a more volatile signal. For example: Older deprivation indices, or prices in a thinner market.
A proxy fallback, or very little underlying data. For example: Prices where there are very few recent sales.
No usable data. The value comes back empty, with the reason. For example: A source is down, or the area is outside our coverage.
Categories that are inferred rather than directly measured are capped at medium confidence by design. And when we watch an area for change, a move backed by too little data is held back rather than flagged, so a one-off blip never triggers a false alarm.
09 · Reproducibility
Every response is stamped with the engine version that produced it, so you always know which version a number came from. You can pin a single request to a version, or pin your whole organisation, and get the same numbers back for as long as you need them.
A change big enough that scores could move, and would invalidate numbers you saved under the old version.
An addition, such as a new category or data source, that does not change the numbers you already have.
A small refinement, with scores staying exactly the same.
Pinning your whole organisation is owner-only, because it matters for regulator-facing audits. A single request can still ask for the latest, and older pinned versions stay available for as long as you rely on them.
Version history
Amenity and transport signals now serve from a precomputed store instead of a live query. Transport access points (rail, metro and bus stops) are sourced from NaPTAN, the national public transport access node register, replacing crowd-sourced OpenStreetMap stops with complete official coverage. Amenity category counts come from a committed OpenStreetMap extract. Both are precomputed per small area, so values are stable and no longer depend on a third-party query at request time. Transport access points (stations and bus stops) now come from NaPTAN, the official national register, giving complete and consistent coverage in place of crowd-sourced OpenStreetMap stops. Amenity counts (schools, food and drink, healthcare, shops, parks and leisure) are served from a precomputed OpenStreetMap extract rather than a live query, removing request-time variability. Counts are precomputed per small area across England, Wales and Scotland, matching the previous per-category search radii so scores stay comparable.
Property prices now come from ONS House Price Statistics for Small Areas (HPSSA), a rolling-year median by small area, instead of a self-computed Land Registry window median. This removes thin-sample noise, so a single recent sale no longer swings an area's price or its year-on-year change; transaction volume still comes from HM Land Registry. Property median price and year-on-year change are now sourced from ONS HPSSA (rolling-year median by LSOA), so small areas with few recent sales no longer show volatile figures. Transaction count continues to come from HM Land Registry Price Paid Data. HPSSA is remapped from 2011 to 2021 small-area boundaries to match the current geography (a 2021 split inherits its parent's figure, a merge takes the mean of its parents).
Intent-aware scoring. Every preset now scores the same seven categories (safety & crime, deprivation, property, schools, amenities, transport, environment) using the full source set; intent changes only how those categories are weighted. Seven standard dimensions for every preset (safety & crime, deprivation, property, schools, amenities, transport, environment), replacing the previous five-dimension-per-preset sets. Intent-aware weighting: the same category contributes differently by decision type (property weighs more for investing, transport for business, deprivation for moving). Business and investing presets now score directly from property and schools signals instead of derived demand proxies. Custom weights let you re-weight the same seven categories for any preset.
Deterministic UK area-intelligence engine. Four decision presets, five weighted dimensions each, per-dimension confidence, and full source attribution. Scores are computed from public data by fixed formulas; no AI in the scoring path. Deterministic scoring: the same inputs and engine version always produce the same score, with no AI in the scoring path. Four decision presets (moving, business, investing, research), each weighting the categories for its use case. Area-type benchmarks (urban, suburban, rural) so scores are fair across settlement types. A confidence level and a plain-language reason on every value, so you always know how solid each number is. Seven signal categories from public data: crime, deprivation, property, schools, amenities, transport, environment. Every value carries its source and observed period. Normalised values plus national and regional context for cross-area comparison. The engine version is stamped on every response and can be pinned, so scores stay reproducible for audits.
10 · For your team
Levers are the controls that shape the API for your keys. All four are opt-in, so if you leave them alone everything works exactly as it does by default. Role-based access, white-labelling and IP allowlisting sit alongside them.
Choose which signals your team's keys can see, so everyone works from the same agreed set.
Save a scoring weighting for your organisation and reuse it across the team, so every score is worked out the same way.
Pin your whole organisation to a specific engine version, so every score stays reproducible for as long as you need it.
Define your own set of areas, so similar areas means similar within your portfolio, not the whole country.
11 · Scope and limits
Said up front to save you time. This is decision-grade area screening and analysis. It is not valuation, not lending, and not address-level today.
It does not predict the market value of a specific property. Use a dedicated AVM for that.
It is an input to enrich your own models, never a decision on any individual's affordability or creditworthiness.
We score small areas, not individual addresses. For a specific property, pair us with an address-level source.
An area right on the boundary of another deserves a closer look. These boundaries are administrative, not behavioural.
Deprivation correlates with protected characteristics. In regulated workflows, buyers stay responsible for fair-lending compliance.
Source-backed, placed in context, confidence-rated and version-stamped. Everything a team needs to build on UK area data with confidence.