# How an E&S carrier ranks its submission queue by expected value (https://riska.io/blog/es-carrier-ranks-submissions-by-expected-value)

_2026-10-12 · Riska · use-case_

An E&S carrier writing property and crime uses Riska to score every submission, renewal and task on expected value: premium size, appetite fit and likelihood to bind, lifted by how close the effective date is. The queue is ordered by that score with the reasons beside each rank, so underwriters start on the accounts most worth winning and decline the rest on the first read. People still make every quote, referral and decline.

## Key takeaways

- A queue sorted by arrival time hands the underwriter's best hours to whichever broker emailed first, not to the account most worth winning.
- Riska scores every submission, renewal and task on expected value: premium size, appetite fit and likelihood to bind, lifted by how close the effective date is.
- The likelihood to bind comes from a model trained on the carrier's own quote and bind history, and it says when it does not know rather than guessing.
- Every rank carries the reasons that put it there, and one click opens the rule-by-rule trail behind the appetite score.
- The rank is an order, not a verdict. A person still quotes, refers and declines every submission, and can work the queue in any order they choose.

> **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. Every premium, score and probability below describes the example, not a measured outcome.

## What a first-in, first-out submission queue costs an E&S carrier

Picture a carrier writing excess and surplus property and crime through wholesale brokers. Six underwriters, one shared queue, and on a Tuesday morning eighty-odd submissions waiting. They are sorted the way most queues are sorted: newest at the top, or oldest, depending on who set up the view.

Somewhere around row forty is a $185,000 warehouse renewal from a broker who binds with this carrier more often than not, effective in nine days. Above it sit a dozen submissions that will never bind: a frame building two blocks from the coast, a fleet with a hazmat radius the guide excludes, three duplicates of the same restaurant group from three brokers.

Every one of them gets read before anyone reaches row forty.

Nobody chose that. The queue chose it, by treating every submission as equally worth an underwriter's first hour. In E&S, where the same risk often reaches several markets at once, the account that waits a day is frequently the account someone else quoted.

> **DEFINITION: Expected value of a submission**
> The premium a submission is likely to bring in, weighted by how well it fits appetite and how likely it is to bind if quoted. A $300,000 risk with a one-in-six chance of binding is worth less of today's attention than a $90,000 risk the broker binds two times in three. Urgency is added on top: an effective date inside the next two weeks moves a submission up, because its value expires if nobody acts.

## How a submission earns its place in the queue

Following the warehouse renewal from the broker's email to the top of the list.

1. **Score it against appetite on arrival** As soon as the submission is read, it is scored against the property appetite model the carrier configured: construction, year built, total insured value, protection class, loss experience, CAT exposure. A knockout rule decides the outcome outright. Otherwise the score lands in a band that reads high, refer or out of appetite.
2. **Estimate the likelihood to bind** A model trained on the carrier's own quotes and binds reads the signals that history says matter: the broker and their hit ratio with this carrier, the line, the premium band, whether it is a renewal, how quickly the broker answered last time. It returns a probability, not a label. When the history behind a broker is thin, it says so, and the rank leans on premium and appetite alone.
3. **Weigh premium and urgency** Premium comes from the indicated quote when there is one and from the exposure on the submission when there is not. The days left before the effective date lift the score as they shrink, so a renewal due next week outranks a new-business risk due next quarter.
4. **Order the queue and show why** The queue sorts by expected value, highest first, and every row shows the two or three inputs that moved it most: strong broker history, high appetite, effective in nine days. Click the appetite score and the rule-by-rule trail opens, from the raw value to the weighted points.
5. **Re-rank when something changes** A fresh loss run lands, a broker sends the missing schedule, the TIV is revised. The appetite score moves with its inputs, the likelihood is recalculated, and the submission finds its new place without anyone re-sorting a spreadsheet.

### One Tuesday morning, ranked by expected value

Four submissions from the same queue, in the order Riska puts them and with the reason shown beside each rank.

| Submission | Premium | Appetite | Likelihood to bind | Effective in | Rank and top reason |
| --- | --- | --- | --- | --- | --- |
| Warehouse property renewal | $185,000 | 84 · High | 62% | 9 days | 1 · Strong broker history, due soon |
| Distributor crime and employee dishonesty | $48,000 | 91 · High | 55% | 21 days | 2 · Clean loss runs, high fit |
| Coastal frame habitational property | $310,000 | 52 · Refer | 17% | 45 days | 3 · Large premium, referral on CAT exposure |
| Trucking fleet auto, 40 units | $240,000 | Out of appetite | Not scored | 14 days | 4 · Knockout: hazmat radius outside guide |

_Illustrative figures describing the example team, not measured results._

## Where the underwriter still decides

The rank decides what gets read first. It does not decide anything else.

Nothing is declined because it sank to the bottom. The trucking fleet above still goes to a person, who sees the knockout rule that fired and the note written on it, confirms it, and declines on the first read instead of the fifth. A referral still needs a sign-off. A quote is still priced and bound by someone with the authority to do it.

An underwriter can also take the third row before the first. When they do, the record keeps what was ranked above it at the time, so a manager reviewing the week sees where judgement overrode the order and why, rather than a queue that silently rearranged itself. Those overrides are also the most useful thing the model can learn from.

The honest limit: a likelihood model is only as good as the history behind it. A new program, a new broker or a class the carrier has never written starts with nothing to learn from, and Riska shows the probability as unknown rather than inventing one.

## Ranking the work on the record it is underwritten on

Most carriers who try this build it beside the system: a scoring spreadsheet, a BI dashboard, a list someone exports every morning. The rank goes stale by lunch, and the reasons live somewhere the underwriter is not.

Riska ranks the work where the work happens. [Submission triage](/platform/submission-triage) reads and clears the submission, [appetite scoring](/platform/appetite) records the rule behind every referral and knockout, and the quote is priced and bound in the [underwriting workbench](/platform/workbench) on the same record that carries its rank. The score, the reasons and the decision that followed stay together, which is what lets a team see whether its best hours went to its best accounts.

## Questions we get asked

### Is this a real customer?

No. This is an illustrative use case describing how an E&S carrier writing property and crime would rank its submission queue on Riska. It names no customer and reports no results.

### How does Riska estimate the likelihood that a submission will bind?

A model trained on the carrier's own quote and bind history scores each submission on signals such as the broker's hit ratio, the line, the premium band and whether it is a renewal. It returns a probability, and shows it as unknown when there is too little history to say.

### Does Riska decline low-ranked submissions automatically?

No. A low rank only means a submission is read later. Out-of-appetite submissions arrive with the knockout rule that fired attached, so an underwriter can confirm and decline quickly, but a person makes every decline.

### Can we change how the rank is weighted?

Yes. The appetite model, its weights and its bands are configured by the carrier, and the weight given to premium, appetite fit, likelihood to bind and urgency is set the same way, so the order reflects how your team chooses to spend its time.

### Why not just sort the queue by premium?

Premium alone puts large risks the carrier will never bind at the top. Weighting premium by appetite fit and likelihood to bind moves a smaller account the broker reliably binds above a large one that is out of appetite or shopped to every market.