Plain-English explainers

Notes on data, decisions, and practical AI.

Working notes from the team behind UK Crime Insights — what the data actually says, how to read it well, and where AI helps without overreaching.

Crime data explained

Reading a UK postcode crime report

A walkthrough of every section of a postcode report — what each category count means and how to read the AI summary.

Methodology

Police UK's open data: what's in it, what's not

Where the data comes from, the monthly release cadence, and the limitations every interpretation needs to acknowledge.

Crime data explained

Anti-social behaviour vs public order

The two most-confused categories in UK crime data — what each one actually covers, and why it matters for interpretation.

For estate agents

Crime data in property due diligence

How estate agents can use crime data well — what to surface for buyers, what to omit, and where it adds genuine signal.

For solicitors

Conveyancing and crime data

When crime statistics are material in conveyancing, when they aren't, and how to handle client questions about safety.

Interpretation

The geography of UK crime reporting

Why some areas appear to have more crime than others — and how reporting density distorts apparent geography.

Worked example

Comparing two postcodes: a worked example

How to compare two areas honestly — the right denominator, the right time window, and what to look at beyond totals.

AI methodology

How our AI safety summaries are written

What the LLM step in Crime Insights actually does — and the cases where it can mislead if you read it without the data.

AI methodology

Why we don't predict crime with AI

A deliberate non-feature — what predictive crime models get wrong, and why we describe the past instead.

AI methodology

An AI assistant or a dashboard?

A practical framing for B2B buyers commissioning AI work — when conversational beats visual, and when it doesn't.

AI methodology

Evaluating an AI prototype: a checklist

Eight questions a non-technical stakeholder can ask of any AI prototype to understand what it actually delivers.