The compliance file an inspection actually tests

The same pattern shows up across almost every failure mode worth naming: a roster that says the ratio's fine while breaks leave a room short; vetting or references missing, expired, or completed after someone already started; sleep checks that aren't recorded, or floor-bed supervision that isn't actually maintained; attendance that doesn't match who's physically in the building; staff who can't explain safeguarding, medication, or collection procedures when asked directly; policies that are generic or unimplemented; accident and medication records missing a signature or a follow-up; and a known premises risk with no documented assessment behind it.

The most serious versions of this — safeguarding, supervision, ratio, vetting, collection, fire, and safe-sleep risk — are the ones that matter most, and they're exactly the risks a polished folder can't paper over, because Tusla's inspection process deliberately combines observation, records, and conversation to check whether they agree with each other.

A useful internal test, worth running before an inspection rather than waiting for one: pick five children at random and produce every required record; pick five staff at random and produce vetting, references, qualifications, and FAR evidence; check whether today's roster reconciles against who's actually supervising; check whether child attendance times match the sign-in system; ask a staff member to explain the safeguarding escalation route and the unauthorised-collection procedure without prompting; produce the last fire drill and what followed it; produce the last sleep-check record; and confirm NCS/ECCE claims actually match attendance and real provision.

Bloomffy's Tusla-facing surfaces were built directly against the actual Inspection Notebook's structure rather than a generic compliance checklist, specifically so this test can be run without assembling anything specially — every record type an inspector could plausibly ask for (attendance registers, staff rosters and ratios, the accident book, staff files, fire and insurance records, the policy checklist, complaints, visitor log) is exportable as a clean, filterable, printable report from the same reports hub, on demand, not reconstructed under time pressure.

Retention runs in the background of all of this rather than as a separate manual task: references and vetting records for five years from when someone starts, attendance, medication, and accident/incident records for two years after a child leaves. Bloomffy's leaver-archive design exists for that second number specifically — a frozen, complete record generated once a child leaves, satisfying the retention requirement independent of the live database, with purge requiring explicit admin approval rather than running automatically once the window passes. It's not finished — no screen yet actually generates that archive or handles the purge approval step — but the schema and the retention gate it depends on already exist, which is a more honest place to be than pretending the whole thing is done.