An illustrative hotel group in Miami renews commercial property, general liability and umbrella. The property schedule is four beachfront-area locations worth $83.7M. The question every placement broker asks first is not “who is cheapest” but “who will even look at this” — and the honest answer is in the appetite table, not in memory.
This post follows the market selector: how it ranks every carrier per line against the appetite table, what “strong”, “possible” and “out” mean, why capacity short of the limit asked is a caution rather than a blocker, and how the marketing plan is chosen. Every carrier and figure is invented for the demo book; none of it is market data.
| Actors | Placement broker (adopts the plan), account executive, practice leader on referral |
|---|---|
| Trigger | The submission is approved; Analyze at the marketing stage |
| Lines | Commercial property, general liability, umbrella / excess |
| Agents involved | Market selector, market outreach drafter, placement compliance agent |
| Decisions | Adopt the marketing plan; add or remove a market; whether the surplus lines market is allowed |
| Outcome | A marketing plan with a written reason for every market in it and every market left out |
| What is measured | Markets approached per line, quotes per line, hit ratio by carrier, lines with a single quote |
The appetite table
The market selector ranks against the broker’s carrier appetite table: for each carrier, the lines it writes with maximum capacity and premium range, core or selective appetite, the industries it targets and excludes (by NAICS prefix), the states it writes, its rating and financial size, admitted or surplus, whether the broker is appointed, and the broker’s own results with it — the share of what was sent that bound.
How a carrier is scored
For each line, every active carrier is checked. Some reasons rule a carrier out outright: it doesn’t write the line, it excludes the client’s industry, it doesn’t write in the client’s state, or it is rated below the A- security floor. Everything else adjusts a score and adds a stated reason or caution:
| Reason | Effect |
|---|---|
| Starting score for a carrier not ruled out | 60 |
| Core (not selective) appetite for the line | +10 |
| Targets the client’s industry | +10 |
| The broker’s bound-to-sent ratio with the carrier | +30 × ratio |
| The incumbent on the line | +8 |
| Each caution: selective appetite, capacity short of the limit, premium outside its range, no appointment, surplus without clearance | −12 each |
A carrier with no cautions is strong; with cautions, possible; with a blocker, out. Ranks sort strong before possible before out, then by score. The plan takes the best four per line that are not out, and always keeps the incumbent.
per line"] --> B{"Writes the line?
Industry not excluded?
Writes the state?
Rated A- or better?"} B -- no --> O["Out
with the blocker stated"] B -- yes --> S["Score 60
plus appetite, industry,
results, incumbent"] S --> K{"Any cautions?"} K -- none --> ST["Strong"] K -- some --> PO["Possible
minus 12 per caution"] ST --> PL["Plan: best 4 per line,
incumbent always kept"] PO --> PL PL --> H{"Placement broker
adopts the plan"} style O fill:#fdecec,stroke:#e57373 style H fill:#e8f8f4,stroke:#10b981,stroke-width:2px
The ranking for the hotel group’s property
The incumbent, a surplus lines carrier rated A, is strong: core appetite, capacity up to $100M covers the $83.7M asked. Three admitted carriers are only possible because their maximum capacity — $30M, $50M, $75M — is short of the limit asked. The engine words that as a caution, “a layer, or a lower limit”, not as a blocker: a broker can build a layered program out of capacity that is individually short. A regional admitted carrier whose appetite table lists Midwest and Northeast states was approached anyway and declined for exactly that reason — the declination is logged with its reason, and it counts toward the surplus lines diligent effort later.
Step by step
Per line, every carrier ranked with reasons, cautions and blockers. In one run on another placement the market selector ranked 15 market-lines into a plan and ruled out 34 carrier-lines.
“Adopt a marketing plan: N carriers across 3 lines”, the incumbent kept on each line, every pick with its reasons. The agent never invents an appetite or a rating.
Accepted, or changed to add a market the broker has a relationship with. Either way the plan carries the broker’s name on the decision record.
One submission e-mail per planned market, attaching the approved version — see the outreach post.
If no market has been approached inside 45 days, MARKETING_LATE refers the placement to a practice leader.
What the run looked like
Every agent step is recorded with the model it used (Claude Haiku 5.5 by default for the agents), the tokens, the cost and the time, and the trace shows the three steps: what the placement’s data and the engines say, what the model reasoned, and what was prepared for a person.
What the agents prepared, and what each rests on
| Agent | What it prepared | What it cites | Who decides |
|---|---|---|---|
| Market selector | Adopt the marketing plan per line | Appetite table, the broker’s results, appointments, ratings | Placement broker |
| Outreach drafter | Send the submission to the planned markets (dry run) | The approved submission and the plan | Placement broker |
| Compliance | Note which picks are surplus lines and what diligent effort they will need | Market types, the home state | Compliance / broker |
Edge cases
| What happens | What InsightPlacement does |
|---|---|
| A carrier rated B++ | Out on every line: below the A- floor. It can still be quoted if a broker approaches it, and the comparison flags it hard. |
| A surplus carrier on a placement not cleared for surplus lines | A caution: “needs three admitted declinations first”. |
| A carrier the broker isn’t appointed with | A caution, and a soft flag on any quote; selecting it fires NOT_APPOINTED. |
| New business with no expiring premium | The premium-range check uses the client’s budget if one is set. |
For the account team
You see why each market is in the plan in a sentence you could say to the client, and the incumbent is never dropped by accident.
For broker IT
The appetite table is a pack file the broker owns; the ranking is a pure function with known-answer tests.
For practice leaders
Hit ratio by carrier and quotes per line come out of the same record, so market strategy is reviewed on results rather than recollection.
Questions we get asked
Is the appetite table market data?
No. In the demo every carrier is invented. In production it is the broker’s own appetite knowledge, maintained by the broker.
Can a model reorder the ranking?
No. The ranking is the engine’s. The model explains it; a number in its explanation that is not in the facts throws the explanation away.

