How Streetwise reports are built

What a Streetwise report is — and is not

A Streetwise report assembles publicly available and licensed data about a specific address and its surrounding area: recorded crime, school inspections and results, planning constraints and applications, flood data, sold prices, demographics, broadband, healthcare, transport and local amenities. We fetch the data, compute a small number of clearly defined summary figures, and present it in one place.

It is data, not advice. A Streetwise report is general information. It is not a survey, a valuation, a legal search, a flood risk assessment, or a recommendation to buy, sell or lend. Where we compute a score or a verdict, the derivation is published on this page in full, and the underlying figures are shown alongside it so you can check our arithmetic.

Where a data source fails or a sample is too small to be meaningful, the report says so. We show "could not verify" or "sample too small" rather than guessing, and an unavailable check is never presented as a pass.

Data sources

Every dataset used in a report, what we use it for, and how current it is. Vintages below are the ones actually loaded in the product as of the "last updated" date at the top of this page — not aspirations.

Crown-copyright datasets are used under the Open Government Licence v3.0. OpenStreetMap data is © OpenStreetMap contributors, ODbL.

Crime comparison

Scope: 800 m — a ten-minute walk

The crime verdict counts recorded incidents within 800 m of the property. That distance is a ten-minute walk at the standard 80 m-per-minute planning assumption — the same walkability unit our amenities and transport modules already use. The radius was chosen on that basis alone, before looking at any property's crime data, and it is fixed: it is not adjusted per property or per report.

The "UK average" baseline — how it is derived

The comparison baseline is built in two steps, both approximate and both disclosed:

1. National totals, scaled by area. From ONS 2023/24 national crime totals we derive an approximate six-month incident count for a typical urban/suburban 1-mile (1.6 km) radius: about 450 incidents in total (violence & robbery ≈ 110, burglary ≈ 25, vehicle crime ≈ 40, anti-social & public order ≈ 180, other ≈ 95).
2. Rescaled to the 800 m disc. The area ratio (0.8 / 1.6)² = 0.25 is applied, giving a six-month 800 m baseline of ≈ 113 total incidents (violence 28, burglary 6, vehicle 10, anti-social 45, other 24). The baseline is then scaled linearly to however many months the report covers (months ÷ 6).

Why this is approximate. Scaling national totals by disc area assumes crime is spread uniformly across the country. It is not — city centres are far denser than villages. The baseline is therefore a rough yardstick, and the verdict wording is deliberately directional ("below average", "above average"), never precise. Treat a figure like "29% less than UK average" as an order-of-magnitude signal, not a measurement.

The verdict

The verdict is the ratio of incidents counted within 800 m to the baseline for the same period:

A trend arrow compares the first and second halves of the period; a change of more than ±15% is labelled "rising" or "falling", otherwise "stable".

Known limits of the underlying data: data.police.uk locations are deliberately smoothed to protect privacy (points are snapped to anonymised map locations, so an incident marker is near, not at, the true address), recorded crime undercounts actual crime, and recording practice varies between forces.

Streetwise school scores

The Streetwise school score is a 0–10 blend of Ofsted sub-judgements and DfE performance data. It is not an Ofsted grade, and every school card links to the school's actual Ofsted report so you can read the source.

Step 1 — Ofsted sub-judgements to numbers

Each of the four Ofsted sub-judgements (quality of education, behaviour & attitudes, personal development, leadership & management) is mapped to a number in direct proportion to Ofsted's own four-point scale — evenly spaced, nothing nudged:

A school is only scored when all four sub-judgements are available; otherwise it is shown as Unrated.

Step 2 — performance data to a /10 figure

• Primaries (KS2): the percentage of pupils meeting the expected standard in reading, writing and maths (RWM) is mapped linearly: 40% or below → 0, 100% → 10.
  score = clamp( (RWM% − 40) / 6 , 0, 10 )
• Secondaries (Progress 8): P8 of 0 (national average) → 5; each whole P8 point moves the figure by 3 (−1 → 2, +1 → 8), clamped 0–10. If the school's 95% confidence interval straddles zero, it is statistically indistinguishable from the national average, so it scores exactly 5 — we neither reward nor penalise noise.
  score = clamp( 5 + P8 × 3 , 0, 10 ) // CI straddles 0 → 5
• Secondaries without P8 (fallback): Attainment 8 mapped linearly, 20 → 0, 80 → 10.
  score = clamp( (Att8 − 20) / 6 , 0, 10 )

Step 3 — the weighted blend

If a school has no usable performance data, the score is computed from the four Ofsted components alone with the weights renormalised, and the card is flagged as Ofsted-only (lower confidence).

Step 4 — band labels

The bands are calibrated against the un-inflated mapping so the label a reader sees matches Ofsted's own language: a school rated Good across the board with typical performance composites to roughly 6.5–7.4 and prints "Good" — which is what Ofsted said about it. A Good school with strong results clears 7.5 and prints "Strong"; Outstanding-leaning schools clear 8.5 and print "Excellent".

What is not scored, and staleness

• Independent schools are not given a Streetwise score. Most are inspected by the ISI under a different framework, so their judgements are not comparable with Ofsted sub-grades. They appear on the map, clearly marked, without a score.
• Special schools, FE and other non-standard phases are likewise shown but not scored.
• Stale inspections: where the latest inspection is more than 5 years old, the card is flagged. The flag does not change the score — an old Good is still shown as what Ofsted found — but you should weight it accordingly and check for a more recent inspection.

Planning checks

What we check

Nine constraint datasets from planning.data.gov.uk: Conservation Areas, Article 4 Directions, Green Belt, Areas of Outstanding Natural Beauty, National Parks, Ancient Woodland, Sites of Special Scientific Interest, Listed Buildings, and World Heritage Sites.

How

We query each dataset for geometries within roughly 1 km of the property, then test whether the property's coordinates fall inside each constraint polygon (a point-in-polygon test performed locally). Listed buildings are point records, so we instead measure the distance to the nearest one and flag anything within 100 m. Recent planning applications come from planit.org.uk: applications within 300 m over the last 3 years.

The unknown-state policy

Caveats

• planning.data.gov.uk is compiled from local planning authorities' submissions. Coverage is incomplete and varies by authority — a constraint missing from the dataset can still exist on the ground.
• planit.org.uk aggregates council planning portals; coverage varies by council and may lag real time.
• These checks are not a substitute for a solicitor's local authority search (LLC1/CON29), which remains the authoritative record of charges and constraints affecting a property.

Flood screening

Method

We query the Environment Agency's flood-monitoring API for designated flood areas within 5 km, fetch each area's actual polygon, and test whether the property sits inside one — and if not, how far away the nearest is. Separately, we query the EA's surface-water mapping at the property's exact coordinates (the 1-in-30-year "medium" scenario), measure the distance to river/sea Flood Zones 2 and 3 from the EA's Flood Map for Planning, and pull any live flood warnings and nearby river-level readings.

What the verdict labels mean here

The surface-water band (Very Low to Very High) is reported separately and reflects the EA's modelled surface-water scenario at the point, not our judgement.

What this is not

• This is not a flood risk assessment. "Low" means the property is outside the EA's designated warning and alert areas — it does not mean the property will never flood.
• Some flood mechanisms are not covered or only partially covered by these datasets: groundwater flooding, sewer and drainage failure, burst mains, reservoir breach, and future climate scenarios.
• If an EA source is unavailable when the report runs, that item is shown as unknown with a "check manually" link to the official checker — not as low risk.
• For any flagged property, commission a formal flood risk assessment before exchange, and check insurance terms early: lenders scrutinise addresses inside Flood Warning Areas.

Prices & valuations

Sold prices

Sold prices come from HM Land Registry Price Paid Data — the register of actual completed sales — queried by street and town, with a postcode fallback, and de-duplicated. Land Registry updates monthly, and new sales can take weeks to months to appear after completion, so the most recent sales on a street may not yet be visible.

Street-level medians and indexing to today

The street figure shown is the median of sold prices, which resists distortion by a single outlier sale. To make sales from different years comparable, each past price is indexed to a present-day equivalent by applying cumulative annual rates from the ONS regional House Price Index between the sale year and today. Where a street has enough repeat sales of its own, we blend the regional index with the street's observed trajectory. Two things to know:

• The rates we apply are approximate annual regional figures, refreshed periodically — not the live monthly ONS index. Indexed figures are estimates of what a past sale might fetch today, not appraisals of the specific property, which may have been extended, renovated or neglected since it last sold.
• £/sq ft figures divide the sold price by the floor area recorded on the property's Energy Performance Certificate, where one exists. EPC floor areas are surveyor-recorded at certificate time; bedroom counts derived from EPC data are partly estimated and flagged as such.

The PropertyData cross-check

Where enough property details are available, we also show an automated valuation (AVM) from the PropertyData API, together with its local £/sq ft and rental figures. This is displayed as a separately labelled cross-check — it is a third-party model's output, not blended into our Land Registry-derived figures, and shown so you can see whether two independent methods roughly agree. Neither figure is a valuation for lending or any other formal purpose; only a RICS valuation is.

General limitations

• Data vintage varies by source. Some figures are live (flood warnings, sold-price queries), some are monthly (crime, Land Registry), some are annual (school performance, GP survey, broadband), and some are fixed-vintage (Census 2021, IMD 2025). The table in Section 2 states each one; reports state the period they cover.
• We inherit our sources' errors. If a council under-reports planning data, a police force records differently, or OpenStreetMap is missing a shop, our report reflects that. We surface source names on each module so errors can be traced.
• Scores are simplifications. Every composite figure loses information. The inputs are always shown alongside the score so you can disagree with our weighting.
• Unknowns are shown as unknowns. Failed checks read "could not verify", small samples read "too small to classify". If you see a clean pass, a source actually returned data.
• Streetwise reports are general information, not advice. They are not financial, legal, insurance or surveying advice, and no report constitutes a survey, formal valuation, legal search or flood risk assessment. Before committing to a purchase, verify anything material through the proper channel: a solicitor's searches, a RICS survey and valuation, a formal flood risk assessment, and current school/Ofsted reports.

Corrections log

When we get a method wrong, we correct it and record it here — we do not silently patch it. Methodology changes are dated on this page.

To report a suspected error in any figure or method, email hello@usestreetwise.com with the report postcode and the module concerned. We will investigate, reply, and — where the error is ours — correct it and log it above.