Skip to main content
Underwriting operations · six weeks · spec and UI, solo

Designing a clearance queue when nobody could say what clearance was

The product was named after a process. It had a button for the process. Its dashboard reported on the outcome of the process. Nowhere in the codebase or the spec did anything define what the process actually did — so I wrote the rules, and labeled every line I'd made up.

Open the prototype (opens in a new tab)Demo sign-in — the email is filled in, any password works
  • Complex B2B
  • Workflow and queues
  • Regulated domain
  • Spec writing
  • Design systems
At a glance
Role
Product designer — research, spec, UI, solo
Timeline
Six weeks, 2022
Domain
Commercial insurance underwriting
Status
Specced and handed to engineering

Four minutes end to end. In a hurry? The three findings are the argument — everything else is captures.

To respect client confidentiality, the client and product are unnamed, some interface details are altered, and the figures shown in the prototype are sample data. The decisions, the spec, and the reasoning are mine and are presented as they happened.

Capture A — the screen everything hangs off

The Reptor review workspace: the scanned application on a dark canvas at left, the extracted fields in the center with two low-confidence values flagged for confirmation, and the eight clearance checks running in a rail at right with C1 stopped on a sanctions hit.
The application on the left, what got pulled out of it in the middle, and the eight checks running down the right — the one that stopped the sequence sits at the top. Whatever an underwriter does here, they may have to defend it a year from now.
Verdict first — for people with forty portfolios open

A designer was handed a product named after a process nobody could define, and wrote the definition instead of waiting for one — then labeled every rule he had invented so the client could argue with it.

Screens designed

13

Screens designed

Seven of them rebuilt in round two

Stage flow spec

8

Stage flow spec

Written from scratch, nothing to copy

States modeled

17

States modeled

With owner and reversibility on each

Trade-offs logged

14

Trade-offs logged

Four of them pure business policy

The brief — what this solves, and for whom

A decision that has to survive an examiner, made fifty times before lunch.

The use case

An underwriter with fifty submissions and half a morning.

Submissions arrive as documents, get read by a person, and either enter the pipeline or do not. The decision has to survive an examiner asking about it a year later. The tool they had reported on the outcome of that decision without ever supporting the making of it.

The design problem

The product was named after a process nothing defined.

There was a button for clearance, a dashboard reporting on clearance, and no definition of clearance anywhere in the codebase or the spec. You cannot design the screen for a decision until you know what the decision consists of, so the missing rules became the first deliverable rather than a blocker to report.

The business problem

Time-to-clear is the number being sold.

Every second of reading time on the clearance screen argues against the metric the product is pitched on. But the same screen is the evidence an examiner asks for, and the thing that decides whether an underwriter trusts the machine at all. Only one of those shows up in a demo.

Three findings

Finding 01Severity: high
Confidence 94% — the spec is the evidence

I wrote the client's business rules for them, then labeled every line I'd made up.

Why I flagged this

Clearance was undefined and the client wasn't available to define it. I could design around the gap — a button, a spinner, a result — and wait. Designing around a gap is the safe move and it gets you nowhere.

The risk

Writing the rules myself meant a carrier might go build an underwriting control I'd invented out of market practice instead of their own policy. In insurance that is not a UX problem, it is a regulatory one.

What I did

Wrote the full battery — eight checks, ordered so the regulatory hard stops run before anything expensive — and tagged every rule with its provenance: evidenced in the code, proposed by me, or open, meaning no safe default exists and someone has to decide.

What it cost

Fourteen trade-offs fell out of it, four of them pure business policy that no amount of design closes. Leave a blank in a spec and everyone walks past it. Write down something wrong and somebody finally corrects you.

The ledger — eight checks, in run order

Regulatory hard stops run first

  • C1Sanctions and watchlistEvidenced
  • C2Licensed to write in stateEvidenced
  • C3Broker of record conflictEvidenced
  • C4Appetite — class and hazard gradeProposed
  • C5Capacity against aggregateProposed
  • C6Prior declination in five yearsProposed
  • C7Duplicate submission windowOpen
  • C8Minimum premium thresholdOpen

Evidenced means the rule was already in the code or the documents. Proposed means I wrote it from market practice and the client can overrule it. Open means no safe default exists and somebody has to decide — two of eight, and they are the two that would change the product most.

Finding 02Severity: high
Confidence 90% — it cost reading time and I kept it

The fast version shows one verdict. I shipped the one that shows all eight checks, boring ones included.

Why I flagged this

Clearance ends in cleared or not cleared, and one line of copy would say it. Showing the whole battery is more to read at the exact moment someone wants to move on.

The tension

Time-to-clear is the product's headline metric, so anything that adds reading time argues against the thing being sold. But the same screen is the evidence an examiner asks for, and where an underwriter decides whether the machine is worth trusting.

What I did

Show all eight. Put the stop first and let the clean passes collapse underneath it. The screen an examiner asks for and the screen that earns an underwriter's trust are the same screen — building that twice would have been the actual waste.

What it cost

The screen got heavier and every conversation about it got easier. It also dragged a rule out of a policy document nobody opens: when a submission is blocked because another broker already holds the risk, you never tell the second broker who. The screen handles that now instead of somebody remembering it at 4pm on a Friday.

Capture B — the three ways out

Three outcomes of the clearance battery side by side: a compliance hold where C1 stopped the sequence, a blocked broker-of-record conflict at C3, and the happy path where all eight checks passed.
Three states rather than one success screen and an error. Left: COMPLIANCE_HOLD — C4 through C8 read “not run, sequence aborted”, which is information rather than an omission. Center: a conflict that names the block but never the broker holding it. Right: cleared in 2m 41s, with a 2019 declination carried forward as a flag that follows the submission instead of stopping it.
Finding 03Severity: medium
Confidence 86% — seven rebuilt screens say so

Round one was finished, internally consistent, and the wrong product.

Why I flagged this

Nine columns, five metric cards, a modal for the clearance result, an action menu on every row. Nothing in it was wrong. It was assembled from what a submissions queue looks like rather than from how one is worked, and those two produce similar screenshots and different products.

Evidence

Round one had no way to decide a submission and move to the next one without going back to the list. Every queue tool people actually work in all day has that. I had designed a queue you would return to fifty times a day.

What I did

Went back to shipped tools that do the same job — bulk review, document verification — and rebuilt on what they do. Two-line cells instead of nine columns. Selection raising an action bar instead of per-row menus. The check battery as a live rail rather than a modal you dismiss before you can act.

What it cost

Seven screens thrown away in the middle of a six-week engagement. The rebuild is the reason there is anything to say here about how the queue is worked rather than how it looks.

Capture C — round one

The first-round queue: nine columns, five metric cards across the top, and a per-row action menu on every line.
Finished, consistent, and built from what a queue looks like.

Capture D — round two

The rebuilt queue: two-line cells, an action bar raised by selection, and the check battery as a live rail.
Rebuilt from how one is worked. Decide and advance without returning to the list.
The register — what each choice cost

Every decision bought something and gave something up.

Auto-clear the clean submissions, or make a human touch every one
What I choseAuto-clear only when every check comes back clean and extraction confidence clears the bar. Configurable, and it ships set tight.
What it costsStarting tight makes the headline number look worse in year one. But you can always loosen a rule once you have evidence. You can't un-clear a submission that already went out.
Let underwriters edit the extracted fields, or lock them to the document
What I choseEditable, with the original always one click away and every edit logged. It's the only way somebody fixes bad extraction without leaving the tool, and the only way anybody ever finds out how accurate extraction actually is.
What it costsIt makes the audit model harder. Every edit has to be stored as a correction, not a replacement.
A dense working tool, or the brand's editorial spacing
What I choseDensity, while keeping the brand's type, color and flat square corners. Somebody working thirty submissions before lunch needs rows, not air.
What it costsIt looks less like the marketing site than the brand people would like. I'll defend that one all day — throughput is the job.
Ship without pricing, or hold the launch until the loop closes
What I choseShip it, and design the seam instead of pretending it isn't there: a thirty-second screen that records what got quoted, where, and when.
What it costsIt's a chore, and the person doing it gets nothing back for it today. Skip it and you can't calculate hit ratio, which means nobody ever makes the case for phase two.
Filter in the browser, or on the server
What I choseServer-side search, filtering and pagination from the first release.
What it costsMore backend work before anyone sees anything, to solve a problem nobody's hit yet. The volume question that would settle it is still sitting unanswered.
Open findings — filed against me

What I'd flag if this were someone else's project.

Open · High

None of this has been used in anger.

It comes off a prototype, a spec, and what I know about how the work is done. I specced it and handed it to engineering in 2022. What you have read is the design and the reasoning, not a measured result.

Open · Medium

A polished prototype buys the wrong feedback.

Show people something finished and they give you opinions about the finish. The round-two screens are convincing enough that the rules underneath them stopped getting argued with, which is the opposite of what the provenance tags were for.

Open · Won't fix

The win score is still a black box.

Every submission carries one and nobody could tell me how it is calculated. I designed around it rather than through it. A number an underwriter cannot interrogate is a number they will eventually learn to ignore.