For B2B integrators
Deterministic signals at LSOA grain. Configurable composite scoring with a pinnable engine version. Portfolio monitoring with sample- size-gated change alerts. A typed query plane where AI emits the plan and the database produces the answer. One API key. Five buyer workflows.
Most buyers integrate the API directly into their existing product or underwriting flow. Some prefer to configure how the API behaves from a dashboard. Either way you reach the same deterministic engine and the same data layer.
Your product, underwriting engine, or analytics notebook calls OneGoodArea directly. We are infrastructure underneath your workflow.
Bearer token · oga_ prefix · /v1/
Configure how the API behaves per organisation. Pin a methodology version, save scoring profiles, manage signal bundles, monitor portfolios, register webhooks, control IP allowlists.
Dashboard · multi-tenant · Levers
Listing portals, valuation tools, agent CRMs, search products. The audience already has the buyer or renter on the page; the area context is what comes next.
The problem
Your product surfaces UK properties to users who immediately want to know what the area is like. Building that yourself means stitching crime, deprivation, prices, transport, amenities and schools across mismatched government APIs, normalising indices across England, Wales and Scotland, and reconciling 2011 versus 2021 boundaries. That is a data team you do not want to hire.
Why OneGoodArea
Signals is exactly that data layer, already stitched together. One typed request to /v1/area returns the seven-category catalog at LSOA grain with national-within-country percentiles, normalised positions, per-signal confidence, and source attribution. Scores compresses the catalog into a single 0-to-100 number per audience (one of four scoring profiles). Intelligence handles natural-language search and similar-areas tiles in one call.
Their value
Weeks of integration replaced by one API key. Your area-detail screen ships with comparable percentiles instead of raw numbers that mean different things in Cardiff and Manchester. Same postcode plus profile gives the same score across deploys, so cached UI states stay coherent.
Products they reach for
Drop one endpoint into your property detail page and ship richer area context than your competitor's roadmap.
Carriers, MGAs, and InsureTech platforms that need defensible, dated area inputs for underwriting and continuous portfolio monitoring.
The problem
Underwriting needs deterministic, addressable, dated values you can pin a price to and re-derive on audit. Once the book is written it drifts continuously: prices move, deprivation shifts, crime rebalances. Pricing teams need to know when a tracked LSOA has actually changed, not at renewal but ongoing, and the alert has to be auditable rather than a black-box score.
Why OneGoodArea
Scores returns every dimension's raw score, weight, and per-dimension confidence so the actuary sees the components. Custom weights or a saved org profile (preset_id) lets the carrier lock the model. Monitor saves the insured-location book as a portfolio and detects material moves on demand, with sample-size gating so a 47-percent swing on two sales never earns an alert. Material moves fire signed webhooks. Intelligence find_peers gives a stable peer set; find_insights ranks LSOAs by peer-relative anomaly z-score.
Their value
Reproducible inputs you can defend to a regulator. Per-signal confidence flows into your decline-or-refer logic without you inventing it. Exposure drift detected continuously, not at renewal. Every alert is auditable: raw values, periods, threshold, sample gate, all in the same envelope.
Products they reach for
Configurable composite scoring the actuary can audit, plus HMAC-signed alerts the day a tracked LSOA moves materially.
Residential and commercial lenders whose underwriting models need versioned, pinnable, auditable area inputs that hold up at a model risk committee.
The problem
A regulated lender's model risk register treats every API the underwriting model depends on as a model input. Auditors ask which version of the area score produced this decision, and whether it is byte-equivalent to the score you would compute today. Most vendors version their codebase, not the methodology, so the honest answer is we do not know.
Why OneGoodArea
Every Scores response carries engine_version in the body and X-Engine-Version on the response header. Both can be locked at the org level (methodology pinning, owner-only, write-time validated). The deterministic engine is frozen v2, golden-tested; AI never touches the scoring path. Monitor records every period_from, period_to and pct_change so risk teams can prove they knew on date X. Intelligence echoes the executed plan plus plan_source so any natural-language answer is replayable as a deterministic programmatic call.
Their value
Every decision produced from /v1/score is reproducible to a known methodology version. The model risk register has a 1-to-1 mapping between an API call and the engine that ran. AI-assisted screening is auditable because the LLM only ever emits a typed plan that gets validated and run against deterministic SQL.
Products they reach for
Versioned, pinnable area scoring with a plan-replayable AI seam your model risk committee will sign off.
Retail expansion teams, asset managers, and CRE analytics platforms screening hundreds of UK catchments against compound criteria.
The problem
Picking a site is a ranking problem at portfolio scale. Which areas in this LAD or country meet my thresholds on footfall demand, competition density, transport access, spending power and commercial costs, sorted by which one moved most this year. You do not want a one-area-at-a-time report API; you want to query the universe and rank.
Why OneGoodArea
Signals /v1/areas is single-signal threshold-and-rank within a country or LAD. Intelligence /v1/query is the compound version: up to 8 AND-joined signal filters, sort by any of them, country or LAD scope, capped at 1000 rows. find_peers gives the peer set of your best-performing catchment in one call. Scores with the commercial preset returns the 5 dimensions a site-selection analyst already uses.
Their value
Hundreds of catchment screens compress into one round trip. Criteria become version-controlled JSON instead of a spreadsheet. The shortlist is reproducible: every result echoes the executed plan, so a colleague can paste it back and get the same answer next quarter.
Products they reach for
Screen the whole UK against your compound site criteria in one typed call, then ask for the peer set of your best-performing catchment.
Council planning teams, central-government analytical units, and regen bodies that need defensible, sourced, dated metrics that survive FOI scrutiny.
The problem
Public-sector teams have to defend every number they publish. A black-box AI score is unusable; they need to point at the methodology, the inputs, and the SQL. They also need to compare like-with-like inside a country, not across a methodological border. England's IMD 2025 is not comparable to Scotland's SIMD 2020; pretending otherwise is a methodological lie.
Why OneGoodArea
Every Signal carries an explicit source, observed_period, and confidence_reason. Normalisation is country-scoped on purpose. Scores' research baseline is the balanced default. Monitor produces a lineage-stamped change report (baseline, threshold, sample gate all in the artifact). Intelligence echoes the executed plan so any answer is replayable as a deterministic call. The methodology version is on every response.
Their value
An evidence base that holds up under scrutiny. Country-scoped percentiles instead of false-precision cross-border comparisons. Same methodology version on every report run if you pin it. Honest sample-size gating: the system says when it cannot tell, instead of hallucinating a move.
Products they reach for
Defensible, sourced, dated area metrics with the methodology version stamped on every response. Built for the procurement notice and the FOI request.
These hold regardless of which ICP you fit into. They are why buyers ship OneGoodArea instead of stitching the data themselves.
Every response carries engine_version in the body and X-Engine-Version on the response header. Org-level methodology pinning locks the version per caller, owner-only.
Intelligence echoes the executed plan plus plan_source on every response. Any natural-language answer can be replayed as a deterministic programmatic call. AI never sets the numbers; the database does.
Monitor change detection gates price moves on transaction count (default 8). Static signals produce zero change rows. The system says when it cannot tell instead of hallucinating a move.
Every signal carries source, observed_period, confidence and confidence_reason. fetch_mode is honestly live, store, or hybrid. Lineage stamps (source_snapshot_id, boundary_version) on every persisted row.
Custom signal bundles, saved scoring profiles, methodology pinning, peer cohorts, three-tier RBAC, white-label, per-key IP allowlist. Opt-in and additive on top of the deterministic engine.
Normalisation runs national-within-country. England's IMD, Wales's WIMD, and Scotland's SIMD are different methodologies and we refuse to manufacture a cross-border percentile.
Numbers you can audit. Source-attributed signals, methodology-versioned responses, deterministic engine.
On a 14-case curated corpus measured against claude-sonnet-4-20250514. Published with the methodology.
On every response body and the X-Engine-Version header. Pinnable per org so two runs return the same numbers.
Signals, Scores, Monitor, Intelligence. One API key, one contract, one deterministic engine underneath.
England, Wales, Scotland — three separate percentile spaces by design. No cross-border lies in any score.
Get an API key and integrate against the surfaces that already exist. Pin the methodology version once and your team is on the same numbers every quarter.