Postlens

About Postlens

Postlens answers “what is it like in this UK postcode?” by reading public data and writing short, honest summaries with AI.

What it does

You type a UK postcode. Postlens fetches crime, housing, schools, transport, and environment figures live from public APIs, aggregates them on the server, and asks a language model to write a short overview. Nothing is pre-computed and no personal data is stored.

Data sources

Every figure you see comes from one of these public APIs:

  • postcodes.io — UK postcode lookup and geocoding
  • ONS — Census demographics, housing, income
  • police.uk — Reported crime data (not actual crime totals)
  • Open-Meteo — Air quality (~11km grid estimate)
  • MapTiler — Base map tiles

How Postlens defines the “area”

You enter a full UK postcode (like SM3 8LL), but Postlens analyses the postcode sector around it (SM3 8) — roughly 3,000 households, 10–15 minutes' walk. That is the scale at which crime, housing and amenities give statistically stable signals; a single unit postcode covers ~15 households and would whip-saw with each new month of data.

Each panel labels its scope with a small badge:

  • Area — sector-level. Same answer for every postcode in SM3 8. Most panels are this scope.
  • Block — LSOA-level (~650 households, finer than a sector). Used by IMD and Census when sector aggregation is unavailable; can vary across a sector at gentrification edges.
  • Your address — unit-postcode-level. Used for distance figures like nearest station, where your specific location matters.

The map pin and your full postcode stay visible throughout — Postlens uses the unit only as a precise anchor for distance calculations and as the address you actually typed.

How crime ranking works

Postlens reports crime as a decile (1–10)against the national distribution of LSOAs in England & Wales — the same approach used by Crystalroof, Streetcheck and Plumplot. Decile 1 means the safest 10% of all areas; decile 10 means the riskiest 10%. We deliberately avoid headlines like “X% below the national average”: that comparison mixes apples with oranges (the per-month street-level subset against the full all-categories total) and routinely produces misleading numbers for urban centres.

Two known caveats are surfaced in the panel itself rather than hidden:

  • Greater Manchester postcodes show an amber warning — Greater Manchester Police has under-reported to data.police.uksince 2019, so the displayed decile materially understates real crime. Don't use it to compare GMP areas with elsewhere.
  • Central London commuter cores(City of London, Westminster, Camden, Kensington & Chelsea, Islington, Southwark) show a blue notice — the per-resident rate looks high because day-time population is many times the resident count. Most incidents involve workers, shoppers, nightlife customers rather than residents.

Three layers, each with a clear purpose:

  • Decile (long-term) — population-weighted median across the LSOAs that overlap your sector, refreshed quarterly from the data.police.uk archive.
  • Block range— when LSOAs in the same area differ by 3 deciles or more, Postlens flags the area as “mixed” and shows both ends. Common at gentrification edges and where new builds sit beside older estates.
  • Live monthly trend— the sparkline below the decile is fed straight from data.police.uk's live API, so you still see the most recent month.

AI and its limits

AI summaries are generated by Google Gemini. The model is only allowed to rephrase numbers we hand it; it is not asked to speculate, predict, or invent street names. Every AI response is validated against a schema before it reaches you, and cached so repeat lookups do not re-run the model.

Limits you should know:

  • Crime data is reported crime, not actual crime. Under-reporting varies by category and area.
  • Air quality comes from an ~11 km grid estimate, not a street-level sensor. Treat the numbers as a regional signal.
  • Demographic figures are Census-based and can be several years old.
  • Comparisons to the London average are just that — comparisons, not verdicts.

Who built this

Postlens is built by Enzo Tang. Send feedback via the contact form.