The allowance

Nobody has taken responsibility for the Retail mortgage parameters. (11 exposures)

Either replace them with your own experience and record the basis, or adopt the starting model as your own estimate — which needs a name and a date against it, not a tick. A published study considered and adopted is an estimate; the same study used because nobody looked is not.

The provision matrix for Services has no documented basis for its loss rates.

Record how the rates were derived. An ageing snapshot alone does not establish a loss rate — that requires cohorts tracked to resolution, with collections, write-offs, recoveries and a stated treatment of incomplete outcomes.

117 of 117 loans have no record of how likely they were to default when they were made. So whether their risk has risen since is judged from the rating they started on, against today's economic forecast. A downgraded loan still moves to lifetime losses on that comparison; a worse economy on its own cannot move anything, because today's forecast sits on both sides of it. The 30-days-past-due rule, forbearance and watchlist still apply as normal.

Add a "Lifetime PD at origination" column to the file — the probability of default recorded when each loan was made — and the economic half of the test starts working. Until then, say in the note that an increase in credit risk is judged on arrears, rating changes and forbearance, not on changes in the economic outlook.