How ReadyIQ works

Every figure in ReadyIQ traces back to a real, publicly available Ofsted inspection report. Nothing is guessed, nothing is scraped from anywhere it shouldn't be, and nothing is written or generated by AI.

The process

From PDF to benchmarked intelligence, in three steps.

We read the reports

Ofsted publishes every children's home inspection report as a PDF. ReadyIQ extracts the structured facts from each one — grades, statutory requirements, recommendations, dates, and the inspector's own narrative — the same information anyone could read by hand, just organised.

We benchmark it

Every figure is placed against the right comparison: homes of a similar size, in the same local authority, the same region, and the sector nationally — so a number means something the moment you see it, not just a fact in isolation.

You see it, searchable

Your home's own page, your portfolio if you run more than one, your Responsible Individuals and Registered Managers rolled up across the homes they cover, and the sector-wide patterns behind them all — grade trajectory, requirement severity, distinctive language, narrative insights, regulation frequency, sector consolidation and closures — all in one place, kept current as new reports are published.

No AI, anywhere

Every number is extracted, not generated.

ReadyIQ doesn't run your reports through an AI model and ask it to summarise or judge them. There is no language model anywhere in the data pipeline — not for the figures, not for the benchmarks, and not for the narrative signals described above. Every single one is produced by a documented, plain-English rule that matches specific wording in the report's own text. That matters because a rule either matches or it doesn't — it can't invent a requirement that wasn't there, soften a concern that was, or quietly drift over time the way a model's output can. If a number looks wrong, the rule that produced it can be inspected line by line, and the sentence that triggered it can always be shown.

No AI or language model is used to extract, summarise, or judge any report — every figure comes from a fixed, auditable rule.
Narrative signals — whether a comment about your independent visitor, your IRO, or a child's own words reads as positive or a concern — are tagged by keyword-and-context pattern matching, not an AI's judgement call.
Every tag and figure links back to the exact sentence, in the exact report, that produced it — nothing is a black box.

Methodology, in the open

Nothing in ReadyIQ pretends to be an official Ofsted rating.

Where we go beyond simply reporting what Ofsted wrote — like weighting a requirement's severity, or ranking which phrases sound distinctively "Outstanding" — that's clearly labelled as our own heuristic, with the method behind it explained in plain English wherever it's shown. If we can't show our working, we don't show the number.

Requirement severity is built from two checkable facts: whether a regulation is one of the nine legislatively-defined "quality standards," and how many sub-paragraphs of it a requirement cites — not a subjective opinion about how bad a breach is.
Distinctive-language analysis compares words and phrases that turn up disproportionately often in Outstanding-graded narrative versus other grades — it tells you what tends to appear alongside a good outcome, not what causes one.
Overdue tracking follows Ofsted's actual regulatory timeline — a rolling day-count from your last full inspection — rather than a fixed calendar-year cutoff.
Every benchmark tells you exactly what it's comparing you against, and how many homes are in that comparison group, so a small sample never quietly passes as a reliable trend.