# How a property carrier automates appetite decisions on submissions, referrals and renewals (https://riska.io/blog/property-carrier-appetite-submissions-referrals-renewals)

_2026-10-12 · Riska · use-case_

A middle-market commercial property carrier scores every submission and every renewal against its own appetite model in Riska before an underwriter opens it. In-appetite risks move straight on to quoting, borderline ones stop as referrals with the rule and its note attached, and knockouts go to decline with the value that fired them. Underwriters still own every referral and every change to the model.

## Key takeaways

- An appetite model in Riska is the underwriting guide written as weighted rules, rolled into one score with bands that say what the score means.
- Every submission is scored the moment it enters the lifecycle, and the screening step routes it to quoting, referral or decline from that result.
- A referral carries the rule that raised it, the note written on that rule and the values it saw, so the underwriter starts at the question rather than the file.
- Renewals open on their own a set number of days before expiry and are scored against the appetite model in force today, not the one that bound them.
- Every result keeps the version of the model that produced it, so a decline or a referral can be explained a year later.

> **NOTE: An illustrative use case**
> This post describes how a team of this shape runs the workflow on Riska. It is a composite example, not a report from a named customer, and it quotes no results. Where a number appears, it describes the example's configuration, not a measured outcome.

## What appetite decisions look like at a middle-market property carrier

Picture a carrier writing commercial property for middle-market accounts: a handful of locations each, a schedule of values that runs to a few pages, a loss run and an ACORD 140 attached to the broker's email. On Monday morning one of those arrives for a fourteen-location light manufacturing account.

The appetite guide is a PDF on a shared drive, last revised in the spring. Two underwriters read its line on frame construction near the coast differently. The point at which a risk needs the property manager's sign-off lives in people's heads. And a renewal bound three years ago is about to be quoted again because it renewed last year, even though the carrier pulled back on that county in the meantime.

None of these decisions is hard. Each one is made by hand, one file at a time, by whoever happened to open it.

> **DEFINITION: Appetite model**
> An appetite model is an underwriting guide written as rules a system can score. Each rule turns one fact about the risk, such as construction class, five-year loss ratio or distance to coast, into a score. Rules are grouped and weighted into one overall score, and score bands say what that score means: in appetite, refer or out of appetite. Knockout rules decide the outcome outright, whatever the score.

## How a submission is scored and routed

Following the light manufacturing account from arrival to a decision.

1. **Write the guide as rules** The carrier groups its rules the way its underwriters already think: property risk, occupancy, loss experience and catastrophe exposure. Protection class and broker quality sit alongside as modifiers, which nudge the score inside a narrow range rather than set it.

Each rule picks how a value becomes a score. Tiers for a five-year loss ratio. A curve for total insured value, so a million dollars more moves the score smoothly instead of over a cliff. A fixed score for sprinklered. A pass-through for a wildfire rating that already runs 0 to 100. And every rule says what a blank is worth, so a missing value never passes by accident.
2. **Score it the moment it lands** When the submission enters the lifecycle, every appetite model in scope for that product runs against it. Total insured value comes from the schedule of values and the loss ratio from the loss run. The result is a score, an outcome, and how much each group of rules contributed to it.
3. **Let knockouts decide first** A knockout overrides the score. In this carrier's model, a five-year loss ratio at or above 100% puts the risk out of appetite whatever else it scores. When more than one knockout fires, the most severe one wins, and every one that fired is recorded with the value it saw and the bound it crossed.
4. **Route on the outcome** The screening step in the product's lifecycle carries guards in priority order. In this example a score of 70 or above with no knockout routes to quoting, 40 to 69 routes to referral, and anything below 40 or knocked out routes to decline. Nobody picks the branch by hand.
5. **Send the referral to the desk that can sign it** Entering the referral status starts a workflow that creates a task for the underwriter who holds authority for that kind of risk, linked to the submission. The task opens on the score breakdown and the rule that referred it, not on the inbox.

## Where the underwriter still decides

A referral is not a soft decline, and the platform does not treat it as one. It stops the submission in a status of its own and waits for a person.

What the person gets is the reasoning. The note the underwriting manager wrote on the rule travels with the result, so "refer anything over 40 years old without a roof update" arrives as that sentence, next to the year built it read. If the year built was wrong on the schedule, the underwriter corrects it and the model scores the submission again.

A rule can also raise a requirement: an inspection report, a signed loss control letter. That requirement blocks the quote until someone satisfies it, or waives it with a written reason that stays on the file. Nothing gets skipped quietly.

### Three outcomes, and what each one carries

What happens to a submission in this carrier's setup once its appetite model has run.

| Outcome | What puts it there | What happens next | What the underwriter sees |
| --- | --- | --- | --- |
| In appetite | Score of 70 or above, no knockout fired | Routed straight to quoting in the workbench | The score and each rule group's contribution |
| Refer | Score of 40 to 69, or a referral rule fired | Stops in referral, with a task for the desk that holds authority | The rule that referred it, its note and the values it read |
| Out of appetite | Score below 40, or a knockout fired | Routed to decline, with the reason ready for the broker | The knockout that fired, the value it saw and the bound it crossed |

_Illustrative configuration. The bands, thresholds and routes are the carrier's own choices, not defaults._

## Re-scoring renewals against today's appetite

Appetite moves. This carrier tightened on coastal frame construction in the spring, and accounts bound before that still sit on the book.

So renewals open on their own, 90 days before expiry in this example, as a new term on the same policy. The renewal re-enters the same lifecycle and is scored like a new submission: against the appetite model in force today, with the facts on the renewal. Last year's result does not carry over. An account that bound when the carrier wrote coastal frame freely meets the same knockouts and referral rules as a new one would.

The renewal queue filters on the appetite outcome like any other column. The renewals that now refer or fall out of appetite are a saved view three months ahead, not a surprise on the broker's call two weeks before expiry. A carrier that wants different rules for renewals can apply a model only to renewal terms, for example to go easier on an account with five clean years.

## Appetite on the same record as the decision

Most carriers keep appetite in a guide, the referral in an email thread and the renewal in a diary. Each hand-off is a place where the reason for a decision gets lost.

Riska is built so that [appetite](/platform/appetite) is a model that scores the record the underwriter works, and its outcome is something the lifecycle can route on. [Submission triage](/platform/submission-triage) lands the submission on that record, [workflows](/platform/workflows) carry the referral to the right desk, and every run keeps the rules, the values and the version of the model that produced it. That is what lets a carrier change its appetite on Monday and still explain last October's decline.

## Questions we get asked

### Is this a real customer?

No. This is an illustrative use case describing how a middle-market commercial property carrier would run appetite decisions on Riska. It names no customer and reports no results.

### Can Riska decline a submission without a person?

It can route a submission to a decline status the moment a knockout fires, if the carrier draws its lifecycle that way. Carriers that want a person on every decline route knockouts to a review status instead; either way the knockout and the value that fired it stay on the record.

### What happens when a submission is missing data the appetite model needs?

Each rule declares what a blank is worth: the worst score, a fixed score the carrier chooses, or skipped with the other rules reweighted. A missing value is never treated as a pass by accident.

### Do renewals keep last year's appetite score?

No. A renewal is scored again when it opens, against the appetite model in force at that moment and the facts on the renewal term, so a change in appetite reaches the renewal book before it comes due.

### Can one product run more than one appetite model?

Yes. Every model in scope runs and keeps its own score and outcome, so a property model and a catastrophe model can score the same submission without one overwriting the other.

### Can we see why a submission was referred months later?

Yes. Every run records each rule's score, every knockout that fired with the value it saw, and the version of the model that ran, so a referral or decline can be explained after the model has changed.