The funnel binds twice as much of the worst programme in the book, and it started doing it before the losses arrived
Restated at realistic scale. Counts are the bordereau's own, restated to the $651M book used across this set. The split of the funnel into new business and renewal is modelled on stated parameters and labelled in every citation: the bordereau carries submissions, quotes and policies but not which of them were renewals, and it carries no declination reason at all.
Hit ratio reads on the book and barely moves on any cut — but that number is a blend of two different businesses. Renewals bind at . New business binds at . Read the new-business half on its own and the flat line disappears: it runs from on Golf & Country Club to on Propane & Fuel Dealers. And it runs with the loss outcome, not against it — correlation across the fifteen programmes is .
So what? An MGA that binds twice as much of what it is shown on its worst programme is not selecting risk — it is absorbing what the rest of the market has put down. The quote rate is not where it happens: the book quotes propane at against a book . It happens at conversion. Of the propane quotes issued, are accepted, against across the book. The book is winning propane because it is the cheapest paper in the room, and Brief 01 is the invoice.
what the book binds of what it is newly shown
twice the book, on the programme running 119%
against 19.8% book — the gap is at price, not selection
appetite moves with loss instead of against it
Four numbers the rest of the brief explains. The blended 23.1% is not among them.
New-business hit ratio by programme · June 2026
How much of what each programme is newly shown it ends up writing. The dashed line is the book. Bars are coloured by what that programme cost over the rolling year, not by the bar.
Colour is rolling-year combined ratio, not hit ratio: red is a programme above break-even, green one comfortably below. On a funnel that was expressing appetite, the colours would run the other way.
The four tallest bars in this chart carry rolling-year combined ratios of 119.0%, 97.7%, 102.8% and 103.6%. The four shortest carry 84.6%, 88.5%, 89.8% and 90.7%. The book binds most of what it is shown on the programmes that cost it the most. Two programmes break the pattern and are worth naming: Habitational runs despite a 101.8% year, and WineRe runs on an 88.6% one — a niche where the book is one of few markets. The largest submission flow in the book is Public Entity Package, and it binds of it.
New-business hit ratio against rolling-year combined ratio · 15 programmes
Every programme is one dot. Left-to-right is how much of it the book wins. Bottom-to-top is what it costs. A book expressing appetite slopes down. This one slopes up.
Pearson correlation +0.80 across fifteen pairs. Fifteen points is a small sample and the reading is directional, but the arrangement is not subtle: the best and worst programmes in the book sit at opposite corners.
Golf & Country Club binds of what it sees and costs . Propane binds and costs . A competitive programme is one you lose most of; a distressed one is one you win. This chart is the market telling the MGA what its book is worth, and it has been saying so every month.
What happens to every new-business submission · June 2026
Each bar is 100% of the new-business submissions that programme received, split by how far they got. If appetite were being expressed at underwriting, the bars would differ in the middle. They differ at the end.
The first segment is the same width on both bars — propane is screened out at the same rate as everything else, and the bordereau records no reason for any of it. The third segment is twice as wide on propane.
Propane is quoted at against a book — statistically the same step. It is bound at against . The entire divergence sits in the last step, which is the price step. That is a narrower and far more actionable finding than 'appetite is loose': underwriting is looking at propane as often as anything else and pricing it in a way the market accepts twice as readily. At the other end of the book, Golf & Country Club converts of the quotes it issues.
One monitor, two businesses
The 23.1% that appears on the dashboard is a weighted average of a 94.8% and a 9.3%. It moves when the mix moves, and it moves for no other reason.
Where the book's 3,409 bound policies came from, June 2026
New-business share of policies bound is 34.0%, against 33.8% expected.
| New business | Renewal | Blended | |
|---|---|---|---|
| Submissions / renewals offered | 12,396 | 2,375 | 14,772 |
| Quotes issued | 5,862 | 2,375 | 8,237 |
| Policies bound | 1,158 | 2,251 | 3,409 |
| Hit ratio | 9.3% | 94.8% | 23.1% |
| Quote-to-bind | 19.8% | 94.8% | 41.4% |
Renewals are assumed quoted in full, so renewal hit ratio and retention are the same number. The blended column is what the monitor reports today.
A blended hit ratio is stable because renewal retention is stable, and renewal retention is stable in almost every book that is not in run-off. of the bound policies come from the renewal side, so two thirds of the monitor is measuring inertia. The third that measures decisions is the third nobody reads.
New-business hit ratio · Propane against the book
The book line is flat by construction — it is an average of fifteen programmes. The propane line is not, and the band between them is the share the rest of the market handed over.
Propane's rise is continuous — no single month explains it. Retention on the same programme held at 94.9% throughout, so nothing was leaving either.
Propane's new-business hit ratio rose 45% over twelve months while the book held at . Over the same twelve months its combined ratio went to . Retention held at , so the programme was growing at both ends. This line moved first — it was above the book by July 2025, four months before the loss ratio broke out.
Every KPI behind this brief, with its verdict
Moved and flat metrics together. The flat ones are doing most of the work here — they are what narrows the finding from 'appetite is loose' to 'the price step is loose'.
| Metric | Actual | Against | Δ | Verdict | Reading |
|---|---|---|---|---|---|
| Hit Ratio · blended, book | 23.1% | 23.1% | 0.0 pp | on level | Stable, and stable for the wrong reason |
| Hit Ratio · new business, book | 9.3% | 9.3% | 0.0 pp | on level | The real denominator of underwriting decisions |
| Hit Ratio · new business, Propane | 18.2% | 15.1% | +3.1 pp | breached | Twice the book, on the worst programme |
| Quote Rate · new business, book | 47.3% | 47.3% | 0.0 pp | on level | Rules out a selection failure at intake |
| Quote Rate · new business, Propane | 46.5% | 47.1% | −0.6 pp | on level | Propane is quoted like everything else |
| Quote-to-Bind · new business, Propane | 39.1% | 33.8% | +5.3 pp | breached | Where the divergence is — the price step |
| Quote-to-Bind · new business, book | 19.8% | 19.7% | +0.1 pp | on level | The comparison that makes 39.1% legible |
| Policy Retention · book | 94.8% | 94.6% | +0.2 pp | on level | Two thirds of the blended number, and it never moves |
| Policy Retention · Propane | 94.9% | 94.7% | +0.2 pp | on level | Nothing is leaving the programme either |
| New-business share of policies bound | 34.0% | 33.8% | +0.2 pp | on level | The mix that makes the blend move at all |
| Correlation · hit ratio to combined ratio | +0.80 | 0.00 | strong | structural | Should be negative; is strongly positive |
| New-business submissions never quoted | 6,535 | 6,510 | +25 | on level | 52.7% of submissions, with no reason recorded |
Expected is each metric's own trailing-twelve-month median. The new-business and renewal split is the modelled layer described at the top of this brief; blended figures, quote counts and retention come from 370 unchanged.
Read the flat rows first. Quote rate on Propane is against a book , and retention is against . Intake is not broken, the underwriters are not declining less, and nobody is leaving. Everything that did not move is what proves the one thing that did is a pricing signal.
The monitor everybody watches is two thirds inertia
Blended hit ratio reads and has not moved a tenth of a point in twelve months. It is a weighted average of a renewal retention and a new-business hit ratio, and of the bound policies come from the renewal side.
Why it matters: This is the whole reason the funnel has looked uninformative. Retention is stable in almost any book that is not in run-off, so a blended hit ratio inherits that stability and reports it as health. Splitting the two is a reporting change, not an analytical one — the numbers were always there.
Appetite runs with the loss outcome instead of against it
Across fifteen programmes, new-business hit ratio and rolling-year combined ratio correlate at . Propane binds and costs . Golf binds and costs .
Why it matters: A hit ratio is supposed to be where appetite becomes visible: a book should win less of what it does not want. This one wins more. That is the textbook shape of adverse selection, and it is visible from two bordereaux joined on programme — no pricing model, no actuarial review.
The divergence is entirely at the price step
Propane is quoted at against a book — half a point apart. It is bound at against — twice the book.
Why it matters: This narrows the fix from a referral-authority question to a rate question. If propane were being over-quoted, the answer would be tighter underwriting rules. It is being quoted normally and accepted twice as often, which means the quote is cheap relative to where the market has moved. That is the same conclusion Brief 02 reaches from the rate monitor, arrived at from the opposite direction.
The funnel moved before the loss ratio did
Propane's new-business hit ratio rose from 12.6% in July 2025 to in June 2026, continuously, with no single month accounting for the move. Its combined ratio reached over the same period.
Why it matters: By July 2025 the programme was already binding a third more of its new submissions than the book. That is the earliest reading of the propane problem available anywhere in this set — earlier than the severity in Brief 01, earlier than the cohort walk in Brief 02. A funnel read at programme grain is a leading indicator, and it was there the whole time.
Half the new-business submissions have no recorded outcome
new-business submissions never received a quote — 52.7% of everything seen. Bordereau 370 carries no declination reason, so a risk declined for appetite and a risk lost on turnaround are the same row.
Why it matters: This is the one genuine data gap in the set, and it is the half of the picture that would settle the question. If the propane submissions the book declined look like the ones it wrote, the problem is price. If they look different, the problem is clearance. Adding a reason code is a small change to the feed and it makes the whole funnel readable.
In which month did Propane's new-business hit ratio first pass the book's, and what was its loss ratio that month?
Dates the leading indicator against the thing it led.
How does the loss experience of propane business written in the last twelve months compare with the propane already on the books?
Tests directly whether the business the funnel let in is worse than what it joined.
Which producers placed the propane new business the book bound, and how does that differ from who placed it two years ago?
Shows whether the flow shifted as other markets withdrew.
Across all fifteen programmes, how has quote-to-bind moved in the twelve months before each programme's worst loss month?
Tests whether the pattern seen on propane repeated elsewhere.
On the four months Contractors GL cut rate, what did its quote-to-bind do?
Reads the price step on the one slice Brief 02 already showed was discounted.