Enterprise regulatory web app

From data entry
to risk judgment

Modernizing a fragmented regulatory review process.
I redesigned how state regulators review financial-professional applications: bringing critical evidence, actions, collaboration, and audit history into one task-based experience.

Role
Product designer
Duration
6 months
Domain
Fintech/RegTech
User
State regulators/SEC
Redesigned FINRA regulator task list Legacy Web CRD task list
In a nutshell

The 60-second version

For recruiters, hiring managers, design and product leaders, and anyone reading between meetings.

01 · The user

State regulators (who work for the government)

They check applications from people who want to work in finance (e.g. financial advisors), then approve or reject them.

02 · The problem

One case. Information everywhere.

Regulators jumped across 5–7 pages and external documents just to understand one application.

03 · My role

Redesigned the end-to-end workflow (0–1)

I led the research, strategy, stakeholder alignment, interaction design, prototyping, and testing.

04 · Key move

One decision. One workspace.

I brought context, actions, comments, rationale, and history together in one task-based view.

05 · The outcome

~28% faster decisions. Fewer errors.

Navigation dropped from 5–7 tabs to 2–3, and errors fell below 10% during testing.

01. Context

Accuracy mattered more than speed, but the system made both difficult.

State regulators review financial-professional applications under fixed deadlines. During the Q4–Q1 renewal season, application volume increases by roughly 1.5×.

Regulators are audited. They would rather be slow than wrong.

State Regulator
30 days
processing deadline
10K+
regulators
30 cases
one regulator / day
1.5×
application increase
02. Research & baseline

The system documented activity. Regulators needed help making decisions.

I observed 12 state regulators during moderated, task-based sessions: 7 newer regulators and 5 veterans. That mix separated genuine usability problems from shortcuts experts had learned over time.

45–50 min
per manual review
20+ min
just to gather context
~18%
decision error rate
~40%
mid-task assistance
Core insight

The problem wasn't how much data existed. It was how poorly that data was organized around a decision.

It was like keeping the spices in the bedroom and the vegetables in the garage. Everything was technically available, just never where the cooking happened.

03. Mapping the workflow

The happy path hid a lot of unhappy work.

With the PM and engineering lead, I mapped the system-generated path alongside manual-review edge cases, identifying the intended flow and the touch points where a straightforward review turns into a scavenger hunt.

Regulator action Workflow friction
Step 1: Entry. Regulators enter through a personalized queue built around their assigned work. Flow: the regulator logs into the FINRA regulator workspace, lands on a personalized task dashboard, then sees queue counts by status and priority. No major friction recorded at this step.
Step 2: Discover. Filters surface higher-risk filings and the first moments that require interpretation. Flow: filters the queue by date, status or risk signal, scans the task list for high-risk indicators, then selects a filing to review. Friction: risk signals need interpretation. flags an anomalous disclosure count; a rule code or status appears unfamiliar; the filing shows conflicting information; a required disclosure is missing or unclear.
Step 3: Gather evidence. Case context comes together, but important history still sends regulators elsewhere. Friction: context is split across pages and documents, so history, prior actions and filings have to be pieced together from separate places.
Step 4: Interpret risk. Regulators decode rule language, missing information, and the next action to take. Friction: the system doesn't explain enough, leaving the regulator to interpret terminology and next steps unaided.
Step 5: Investigate deeper. Related filings and cases demand another round of cross-checking. Friction: related work is hard to verify across records.
Step 6: Decide. The final outcome, rationale, and comments are recorded, sometimes in another system, so the reasoning ends up outside the secure environment.
04. Diagnosing the legacy experience

Three screens. Three different kinds of friction.

Instead of treating the legacy flow as one vague usability problem, I separated navigation friction, missing decision context, and gaps in the audit trail.

Step 1: Entry

Familiar for experts, difficult for new users

  • Experienced regulators relied on memory to navigate.
  • High-risk queues were easy to miss.
  • New regulators needed time to learn the layout.
The strategic decision to descope
  • Most users spent only 6–8 seconds here
  • High engineering risk (downstream dependencies)
  • Focused on higher-impact parts of the workflow
The legacy Web CRD site map: seven columns of links under CRD Main, Forms, Individual, Organization, Notifications and Reports, with Manual Approval buried in the Jurisdiction Acceptance Queues list.
Step 2: Task list

Task list showed the task, but not the context

  • Critical context missing
  • Manual cross-checking required
  • No inline assignment of task
  • Forced users to the profile details page
Why it broke
  • Key risk signals were separated from the task.
  • Task assignment created a separate workflow break.
The legacy Manual Approval Queue: a dense grey table of three rows with columns for assign, individual, firm, jurisdiction dates, disclosure counts and a status dropdown.
Step 3: Task details

It recorded the outcome, but not the reasoning

  • The system captured status changes.
  • Comments and rationale lived in spreadsheets, documents or email.
  • Important context could be lost between regulators.
Why it broke
  • ‘What’ changed was stored, but ‘why’ it changed was not.
  • “Source of Truth” lived outside the secure environment (separate spreadsheets & docs)
The legacy Composite Information screen: stacked rows of legal name, employment, affiliation, addresses and disclosure counts in a plain bordered table.
05. Product strategy

Design around the decision, not the database.

I translated the research into four measurable KPIs. Each connected an interface choice to a behavior the product needed to change.

KPI 01

Time to
decision

Put related evidence inside one task view.

KPI 02

Decision
accuracy

Surface history, prior actions & risk signals where decisions happen.

KPI 03

Navigation
depth

Replace page-based detours with a task-based structure.

KPI 04

Mid-task assistance

Make terminology and next steps clearer without slowing experts.

06. Iteration and trade-offs

Keep expert speed. Add novice confidence.

An earlier iteration split task information into two rows. Veteran regulators lost their familiar left-to-right scan, while newer users found the dense context and error states unclear. Eye-tracking told us why.

Rejected two-row task layout
Attention patterns

Same screen. Two different ways of looking for an answer.

Newer regulators searched broadly for orientation. Veterans focused on a smaller set of decision-critical fields. The design needed to support both without making either group slow down.

Heat map showing broad scanning by novice regulators

Novice regulators

Broad, exploratory scanning

Heat map showing selective scanning by veteran regulators

Veteran regulators

Selective, decision-focused scanning

What we learned

The information was there. Scanning it was hard.

01

Familiar pattern broken

Two rows interrupted the scan pattern experienced users relied on.

02

Important details buried

Too much information competed for attention.

03

Errors lacked direction

Users couldn't tell what went wrong or what to do next.

07. Final experience · Task list

Start the day knowing what needs you next.

The queue had to carry urgency, context, and ownership in a single scan: quick enough for a veteran, legible enough for someone in their first month.

Final designs: task list (Queues) page

A decision-ready task list

The queue surfaced urgency, task context, and ownership in one scan. Experts could move quickly; newer regulators could stay oriented.

Core outcome
  • Colour-coded urgency without relying on colour alone
  • Inline assignment instead of a dispatch detour
  • Hover guidance for unfamiliar statuses
Animated walkthrough of the redesigned Manual Approval Queues task list: colour-coded urgency badges, inline assignment, and hover guidance on unfamiliar statuses.
Final designs: task list card

In line context

Cognitive support — balancing experts and novices:

  • Preserved expert scan pattern
  • Added colour-coded urgency
  • Inline hover guidance
Operational efficiency
  • Inline “Assign” action
  • No separate dispatch view
  • Reduced workflow breaks
A single task row from the redesigned queue, with a hover tooltip explaining the No Exam for RA Position reason and an inline Assign To panel listing colleagues to hand the task to.
07. Final experience · Task details

One workspace for context, action, collaboration, and audit.

Once a regulator opened a task, everything needed to decide, and to defend that decision in an audit later, had to sit in one place.

Task details page: profile, risk & context

See the person first. Then understand the risk.

  • Key details came first.
  • Disclosures were grouped clearly.
  • Risk signals stood out.
  • Context stayed visible.
Core outcome
  • Faster, more confident reviews.
Individual Review Worksheet: the applicant's identifiers, address history and firm association on the left, a disclosures table in the middle, and rep details, filings and actions in a right-hand rail.
Task details page: unified task space

The decision happened where the evidence lived.

  • Decisions happened in context.
  • Comments stayed with the task.
  • Every change was recorded.
  • Related tasks appeared together.
Core outcome
  • Fewer errors and less repeated work
Task details page: Task aggregation & Batch processing

All open tasks related to that profile, surfaced together.

  • All open tasks surfaced together
  • Profile reviewed once
  • Batch adjudication enabled
Core outcome
  • Absorb context once.
  • Resolve multiple decisions.
Task details page showing an open Manual Approval task with its review statuses and registration status, and beneath it a list of the other open tasks for the same individual, so related tasks can be adjudicated together.
07. Final experience · Session timeouts

Preventing surprise session timeouts.

The redesign brought each decision into one task view. Keeping regulators in that view meant designing the whole workflow around it, including the moment a session is about to end. FINRA’s 30-minute inactivity rule stayed exactly as it was; what changed was what the interface told regulators before it applied.

Before · the existing timeout

No warning before the logout

Idle sessions already ended at 30 minutes, with no sign of the time left. Regulators came back from a PDF or an email to a login screen and MFA.

The real gap

The rule was in place. Timely feedback wasn’t.

The CRD login screen with a small pink 'Session timed-out' banner in the corner. It is the first sign the session has ended; nothing warned the regulator beforehand.
Engineering constraints

What keeps a session active, and what doesn’t

Keeps the session active

  • Another tab, same app
  • Continue sessionOnce the server confirms

Doesn’t count; the clock keeps running

  • Opening or reading a PDF
  • Another FINRA appSame login
  • Same app, different login
  • Email or other appsOutside FINRA
After · in-app warning

One warning, three minutes out

At three minutes left, a modal explains why, counts down, and offers Continue session or Log out now.

Optional opt-in

The desktop notification checkbox. Implementation status unverified.

The Manual Approval Queues page dimmed behind a modal reading 'You will be logged out soon', explaining the 30-minute inactivity rule, with a 02:59 countdown, an Enable desktop notification checkbox, and Log out now and Continue session buttons.
Desktop notification

A nudge at three minutes, and again at one

For regulators working in another app, a desktop notification appears at 3 minutes and again at 1 minute remaining.

Browser support

We started by making sure it worked in Chrome, Safari and Internet Explorer, the browsers about 90% of regulators used.

Concept mockup of Outlook on Windows with a notification in the corner from Regulator Workspace via Chrome: 'Your session ends in 3 minutes. Return to Regulator Workspace to continue.'
Failure and expiry states

No renewal until the server confirms

Every path waits for the server. A failed renewal can be retried while time remains; a late one routes to sign-in.

Modal reading 'We couldn't confirm your session extension. Check your connection and try again.' with the 02:59 countdown still running and Log out now and Continue session buttons.
Renewal failedTime remains; try again
Modal over the login screen reading 'Your session has ended', explaining the 30-minute inactivity rule, with a single Log in to continue button.
Session expiredRoute to sign in
Timeout warning · one-day beta
11of 12
regulators renewed their session before the inactivity timeout

Moderated testing with the same 12 regulators, then a one-day beta. A small sample, not a product-wide rate.

08. Impact

Better decisions and fewer operational workarounds.

Pre-launch results came from moderated testing with the same 12 regulators. After launch, the team monitored the first renewal season to understand the business effect.

Pre-launch · moderated testing
28%
Faster decisionsPut related evidence inside one task view.
<10%
Decision errorsDown from the ~18% baseline.
2–3tabs
Tabs per reviewReduced from 5–7 tabs.
25%
Mid-task assistanceDown from ~40% for newer regulators.
Post-launch · business impact, year 1
↓30%
Drop in support ticketsMore guidance inside the workflow meant fewer “How do I do this?” requests.
↓37%
Lower audit kickback ratePutting evidence beside the action reduced errors in live filings.
09. Reflection

The biggest win wasn't showing more information. It was making the right evidence impossible to miss.

Enterprise modernization is not a visual-cleanup exercise. It is a prioritization problem: understand where decisions break down, choose the highest-leverage surface, and connect every change to an observable outcome.

01

Auditability is UX

Capture rationale where the work actually happens.

02

Experts and novices coexist

Preserve scan speed; layer guidance on top.

03

De-scoping is design work

The right “not now” protected the outcome.

Press Esc or click outside the image to close