Reinsurance took thirty points off the month, and the layer that did it is running at 191%
Restated at realistic scale. Every ratio is computed from the bordereaux unchanged; the levels are restated to a $651M annual book with severity set per line of business. The reinsurance layer and the individual losses are modelled on stated parameters and labelled in every citation.
Propane & Fuel Dealers ran a gross combined ratio of in June. Claim count was — unchanged — while severity reached . Catastrophe claims made up of the month against a 2.9% normal, and reserve ratio moved to . Five losses breached the $250,000 per-risk retention, so the month lands at net. Pricing had nothing to do with it: average rate held at .
So what? The net result looks survivable because the reinsurance layer absorbed it — and that layer has now paid . A layer running at nearly twice its premium does not renew on the same terms. On 1 January the retention goes up or the rate does, and the net result moves toward the gross . The exposure is not this month. It is the renewal.
fifty points above the programme's own normal
after $1.195M of recoveries, and still over break-even
the entire move, on an unchanged claim count
the reinsurance layer that absorbed it is now loss-making
Four numbers the rest of the brief explains.
Combined Ratio · distance from each programme's own expected level
Distance from a programme's own trailing-twelve-month median matters more than the raw ratio. Right of the line is worse. The verdict on the right still reads the actual ratio.
Bar length is distance from that programme's own expected level, on one shared scale spanning ±55 points. Habitational at 104.5% is above break-even but only 5.7 points above its own normal; Municipal Law Enforcement at 100.1% is below its.
Fourteen programmes finish within six points of their own level. Propane is fifty points out at . Above break-even and above normal are different findings — Commercial Trucking at 103.0% is the first and not the second, and needs a pricing conversation rather than an incident review.
Incurred loss by class and loss type
The next shared dimension down. Both classes are over — but not for the same reason, and not by the same amount.
Both bars are on one shared scale, so the lengths compare directly. Energy carries 48% of the programme's earned premium and 57% of its incurred loss.
Energy at is the obvious half. But Transportation at is the finding people miss: the delivery fleet is thirty points over its own normal too, and catastrophe is 43% of its loss. A risk-engineering programme scoped to storage sites would leave half the problem on the road.
Combined Ratio by state · Propane & Fuel Dealers, June 2026
The programme is written in four states. Every other state is unshaded because there is no business there — a thin map is coverage, not absence of risk.
Hover or tab to a state for its loss ratio, premium, claim count, severity and ceded amount.
Frequency against severity · all 15 programmes
Loss cost is frequency times severity, and the two need opposite remedies. Where a programme sits on this plane is the diagnostic. Bubble size is earned premium.
The dashed circle is where Propane was expected to sit. The arrow is vertical — frequency did not move at all.
The arrow points straight up. Propane did not move along the frequency axis, it moved up the severity axis: at an unchanged . A frequency problem is fixed at the point of quote; a severity problem is fixed with retention, rate and risk engineering.
And these are the five losses that did it
The dark band is what the MGA and the carrier keep; the green band is what the reinsurers pay.
The five sum to $2,445,000 gross and $1,195,000 ceded; with the 21 attritional claims this reconciles to the $2,955,000 in the loss run.
Read what these actually are. Two delivery-vehicle accidents, one customer-site fire after a service visit, one carbon-monoxide injury from an installation, one overfill at a plant. Not one of them is a catastrophic facility event. Every single one arises from something this business does every day — trucks on the road and technicians at customer premises. That is why both classes are over, and it is why this is an operational exposure rather than bad luck.
Every KPI on the month, with its verdict
Moved and flat metrics appear together, so the ruled-out explanations stay visible rather than hidden. Four of the thirteen are here precisely because they did nothing.
| Metric | Actual | Expected | Δ | Verdict | Reading |
|---|---|---|---|---|---|
| Combined Ratio · gross | 144.4% | 93.9% | +53.8% | outbreak | The headline. Fifty points above the programme's own normal, and 24 above the cancellation level. |
| Combined Ratio · net | 114.2% | 100% | +14.2% | critical | What the MGA actually carries after recoveries. Still over break-even. |
| Claim Severity | $113,557 | $64,000 | +77.4% | driver | The entire move. Each claim cost three-quarters again what it usually does. |
| Reported Claim Count | 26 | 26 | flat | rule-out | Unchanged. Appetite and underwriting selection are ruled out as causes. |
| Cat Claim Mix | 22.9% | 2.9% | +690% | precursor | Claims flagged event-related, arriving as one cluster. Visible in 372 before the loss run moved. |
| Reserve Ratio | 45.9% | 35.0% | +31.1% | watch | Reserves strengthened with the severity, so the reported figure understates the ultimate. |
| Paid Loss Ratio | 60.3% | 111.4% | −45.8% | watch | Only 54% of the incurred loss is cash out the door. The rest is still an estimate. |
| Claim Closure Rate | 52.4% | 62.0% | −15.5% | watch | Claims are closing slower. A backlog is building behind the severity. |
| Average Rate | 3.734 | 3.714 | +0.5% | rule-out | Pricing held. This is not an under-pricing story, which narrows the remedy considerably. |
| Litigation Rate | 18.6% | 18.0% | +3.3% | rule-out | Flat. Social inflation is ruled out — this is physical loss, not legal escalation. |
| Policy Retention Rate | 94.9% | 94.5% | +0.4% | rule-out | Nobody is leaving. The book is not shrinking into a worse mix. |
| Bordereau Reject Rate | 2.1% | 3.2% | −34.4% | trust | Cleaner than its own normal, so the loss figures can be believed. |
| Layer Loss Ratio · 12m | 191% | 100% | +91.0% | outbreak | The reinsurance layer has paid $1.91 for every $1 of premium. This is the renewal risk. |
Expected is each metric's own trailing-twelve-month median for this programme. Verdicts: outbreak = the event; driver = what moved it; precursor = fired first; watch = the figure is still moving; rule-out = checked and clean; trust = whether the data can be believed.
The rule-outs do most of the work here. Rate held at , litigation was flat, retention held, and the feed is clean at . Four independent explanations eliminated leaves a physical severity event in one class, which is a much narrower thing to act on than a bad combined ratio.
The same slice, read in order across all three bordereaux and the reinsurance layer
Pricing first, then claims as they emerge, then the loss run, then who ends up paying. Every step is the same programme, the same month — joined on the shared backbone, not on an ID.
| Step | Source | Actual | Expected | Δ | Verdict | Read |
|---|---|---|---|---|---|---|
| Premium · Average Rate | 370 | 3.734 | 3.714 | +0.5% | rule-out | Rate held. The business was priced as intended, so nothing upstream explains the month. |
| Claims · Cat Claim Mix | 372 | 22.9% | 2.9% | +690% | precursor | A cluster of event-coded claims lands in the claims feed first. This is where the month became visible. |
| Claims · Fire share of causes | 372 | 26.8% | 16.9% | +58.6% | context | Small fires rose across the attritional book too — consistent with storm-damaged regulators and appliances, not a plant event. |
| Loss run · Claim Severity | 371 | $113,557 | $64,000 | +77.4% | outcome | The claims feed signal arrives in the loss run as cost per claim, on unchanged volume. |
| Loss run · Reserve Ratio | 371 | 45.9% | 35.0% | +31.1% | amplifier | Reserves strengthen alongside, so the reported outcome understates where this lands. |
| Loss run · Combined Ratio | 371 | 144.4% | 93.9% | +53.8% | scoreboard | The number the carrier reads, and the one that decides whether the programme is retained or cancelled. |
| Layer · Recoveries | Layer | $1.195M | $0 | 5 losses | absorber | Five losses clear the $250K retention and the reinsurers pay, bringing the month to 114.2% net. |
| Layer · Layer Loss Ratio | Layer | 191% | 100% | +91.0% | renewal risk | The absorber itself is now loss-making. That is what converts a bad month into a structural problem. |
Datasets are anchored on different events — written date (370), report date (372), valuation date (371) — so “June” is three different events, not the same policies. The alignment is a correlation on the slice, strong enough to act on, not strong enough to call proof.
Step 2 is the one that matters operationally. The catastrophe signal is in 372 at while the loss run still looks ordinary — the claims feed sees this before the scoreboard does. That is the lead time the whole value pack exists to buy.
The net number is fine, and that is exactly the problem
Gross , net . Reinsurance absorbed thirty points. But the layer has paid over the year and has been hit in four of the last twelve months.
Why it matters: A layer is priced on the expectation that it attaches occasionally. This one attaches every quarter. At renewal the reinsurer lifts the retention, raises the rate, or declines — and in all three cases the net result moves toward the gross. The 144% is not a bad month that reinsurance solved; it is a preview of the net result after January.
Same claims — and the severity came from the fleet and the technicians
Claim count against an identical expectation; severity . Event-coded claims reached , arriving as one cluster. The five losses over retention were two delivery-vehicle accidents, a customer-site fire after a service visit, a carbon-monoxide injury from an installation, and one overfill at a plant.
Why it matters: None of these is a rare catastrophic event — they are the ordinary hazards of running trucks and sending technicians to customer premises, which this business does every day. A weather week turned routine exposure into five large losses. That points the remedy at fleet telematics, driver qualification and installation QA, not at a bulk-plant survey programme — and it means this recurs unless the operations change.
Transportation is over too, and it is the half that gets missed
Energy ran , which is the expected half of the story. Transportation ran — thirty points above its own normal, with catastrophe at 43% of its incurred loss.
Why it matters: A per-risk review scoped to bulk plants and storage sites covers 57% of the loss and leaves the delivery fleet untouched. The Erie rollover with product release is already in the five. Scoping the remediation by class rather than by the headline is what stops half the exposure surviving the fix.
Four independent explanations were checked and eliminated
Average rate held at , litigation rate was flat at 18.6% against 18.0%, policy retention held at 94.9%, and the bordereau reject rate came in at — cleaner than its own normal.
Why it matters: On an incident month the calm metrics do most of the explanatory work. Ruling out under-pricing, social inflation, adverse selection and a broken data feed is what turns a 144% into a physical severity event in one class — a much narrower thing to take to a carrier than a bad combined ratio.
The carrier writes a clean programme of the same size
The relationship reads over the year. Propane is $32.6M of it at 119.0% gross; Pest Control is $33.0M at 90.9%.
Why it matters: This changes what the MGA walks into the renewal asking for. A carrier looking at one bad programme cancels it. A carrier looking at a relationship where half the premium performs has a reason to restructure instead — a higher retention on propane, a rate filing, a risk-engineering commitment. The clean programme is the negotiating position, and it only appears if you read at carrier grain.
How many losses in the last 24 months fell between $250,000 and $500,000?
That band is exactly what the layer absorbed and a higher retention would not have. It is the number the reinsurer will open the renewal with.
How many of the last two years' large losses involved a vehicle or a service visit?
Separates the fleet exposure from the site exposure with a count rather than one month.
Which producers placed the accounts behind the five losses over retention?
Producer is untouched in this brief, and it is where remediation actually gets assigned.
Did event-coded claims cluster in 372 before the loss run moved, in every shock month?
Tests whether the precursor is a repeatable trigger or hindsight on a single month.
How have the reserves set in each earlier shock month developed since?
Tells you whether this programme's reserve-strengthening warning has been proved right before.
Which months in the last 24 did this programme run over 120% gross?
Establishes whether four shocks in twelve months is the pattern or the exception.