Method
From a frozen extract to a pack the chair will actually open
This page is the working sequence we use on Claims Trend Reviews and, in shorter form, on fraud briefings and cohort reads. It is not a mystery process. It is a list of dates and arguments.
1. Name the book
We will not mix private motor with commercial motor, or household with landlord, to make a prettier volume chart. You name the product, the accident years or weather years, and the committee the pack is for. If the chair always asks about average cost per claim, we will still show it — then put mix and lag beside it so the average cannot hide a bulge of light files.
2. Freeze the extract
A usable extract holds claim identifiers, collision or loss date, first notification date, first payment date, cause or peril, payments, opening and current reserve, repairer or contractor, solicitor flag, and reopen indicator. Personal names are not required. If collision date has been copied from notification date, we stop and say so; otherwise every lag chart is theatre.
3. Send a gap list
Within five working days we return a list of fields that are empty, defaulted, or free-text we cannot safely group (garage names are the usual offender). The engagement clock pauses until you tell us whether we should clean, exclude, or sample.
4. Read a sample with claims
Charts without files are how reopen spikes become panic. We agree a sample — often twenty files for a leakage question, fewer for a cluster — and sit with the claims owner. The sample is not random closed-only. Open files that will hit the next ledger belong in the room.
5. Draw few charts
Volume by loss week, the same files by notification week, paid amount by payment week, mix, and reopens with a reason where you have one. For fraud monitoring we replace the heat map with a dated watch list: pairings, counts, and the files on the edge. For storms we prefer a table of days-after-landfall by trade.
6. Write in sentences
Each chart earns a paragraph. If we cannot write the paragraph, the chart does not go in the pack. The fraud annex says what the extract cannot prove. The pack is sent three working days before the briefing so nobody is seeing the lag chart for the first time under fluorescent light.