Claims Severity & Reserve Early-Warning · Brief 01
Claims Severity Command Centre
Every program on one ladder · scored monthly on the large-loss severity tail, not just the headline loss ratio
The one line you configure
Which programs carry a real large-loss tail — and is any of them being fed by discrete severity shocks the attritional book hides?
In a nutshell
Almost the whole book is attritional — nearly all programs sit under a 10% large-loss ratio. One is different in kind: Propane & Fuel Dealers runs a 22.5% large-loss ratio, and on its Energy-class slice in PA · OH · TX it spikes to a ~$120K claim severity and a 341% combined ratio at the shock months. This is fire / explosion tail — a handful of events — not a frequency trend, and it strengthens reserves each time it hits.
At a glance
Every program ranked by large-loss ratio (large losses ÷ incurred; the book sits ~6%)
What moved
What's happening · severity sits calm at ~$25K then shocks to $120K on the driver slice
The slice is quiet at the book severity for most months, then a fire/explosion month drives it clean off the top of the band — a discrete tail, not a rising attritional trend. The shocks are escalating: $68K → $94K → $120K.
Why · the heat is the Energy class in three states — everything else is in-band
Cells are the combined ratio. Read down: only the Energy class is hot. Read across: it burns in PA, OH and TX (all ~132%) and cools to ~98% in FL and CA — the tail lives on the propane risks in three states, not across the footprint.
Where · the tail concentrates in the PA · OH · TX energy corridor
Pennsylvania, Ohio and Texas carry the fire/explosion tail (large-loss ratio ~23–24% vs a ~6% book); North Carolina is milder and the Energy risks elsewhere sit in-band. The action is a per-risk cap and risk-engineering audit on three states, not a book-wide reserve strengthening.
Fire / explosion loss events — PA · OH · TX energy risks (recurring, culminating Jul-26)
A run of fire, explosion and pollution-spill claims on Energy-class propane risks in three states lifts severity from a ~$25K calm baseline to ~$120K at the shock months — a man-made catastrophe pattern that recurs on the Ascot-backed slice, each time strengthening case reserves well before it settles.
So what
nearly all programs are attritional — routine, reservable, in-trend. One is not: Propane & Fuel Dealers carries a genuine large-loss tail, and on its Energy-class PA/OH/TX slice a handful of fire/explosion events push severity to ~$120K and the combined ratio to 341% at the shock months. Because it's tail — not frequency — a monthly loss-ratio glance smooths it away; the severity and reserve view catches each shock as it hits, in time for a risk-engineering audit and a per-risk cap review with the carrier before the treaty true-up.
The points that matter
It's tail severity, not frequency
Claim frequency on the Propane slice is flat at the book rate — the entire move is a handful of fire/explosion/pollution-spill events lifting average severity from ~$25K to ~$120K.
4.8×severity, not countEach shock strengthens reserves before it settles
The reserve ratio jumps from ~40% to 64% at the shock months — the loss is booked into case reserves early, so the severity view flags it well ahead of paid development.
→ 64%reserve ratioThe exposure is contained to one program × class × three states
Energy-class propane risks in PA/OH/TX carry the whole tail; a per-risk cap and a fire/explosion risk-engineering audit with the carrier target it directly.
3 statessurgicalQuestions it already answers
Is the ~$120K a real severity shock or a data artefact of a few large claims?
Real tail — the large-loss ratio moves in lockstep to 55% and cat claim mix to 65% at the shock month, so it's a cluster of fire/explosion events, not a mis-keyed record.
Is this a frequency problem I should re-underwrite, or a severity cap problem?
Severity — frequency holds at the book rate; the lever is a per-risk cap and risk-engineering on the PA/OH/TX energy risks, not tighter appetite on volume.
Which other programs share a fire/explosion peril and could show the same tail?
New Energy Risk is the nearest adjacency on peril but runs attritional today (large-loss ratio ~4%) — worth a watch-flag, not action.