Maxine Zhou About

Launchpad

Improving underwriting efficiency through behaviour-driven design

Turning user data into a smarter navigation model + dynamic viewport to provide a better decision environment.

Context

Overview

My Role

Product Designer (Sole Designer)

Responsibilities

  • Next version of this product
  • UX analysis & insight synthesis
  • Interaction design
  • Information architecture
  • Stakeholder alignment

The Product

An internal underwriting application for insurance teams to review and triage submissions, manage workflows, and make faster and more confident decisions.

Usage at a glance

(In a typical month)

2,720Review sessions per month
224hTotal active usage
26minAverage session length
18Teams actively using

Two pages, one workflow

Underwriters work in a high-volume, high-stakes environment where speed and accuracy are essential. The application's workflow centered around two primary pages:

Pipeline View

A table-based workspace for filtering, triaging and updating submissions.

Pipeline view showing a table of insurance submissions with filters and grouping
Pipeline View: the triage workspace.

Submission Detail

A deep-dive view with full context, documents, analysis and activity history.

Submission detail page showing summary, ranking and activity panels
Submission Detail: the deep-dive view.
Research

Start at the UX audit

As the sole product designer on the next version, I began with a structured UX audit to find where the interface was costing people time.

UX audit findings clustered into themes across the product screens
Identify and prioritise UX issues by conducting UX Audit.

Understanding real usage

I analysed funnel behaviour, heat maps, feature usage and session data to understand how underwriters actually worked.

Questions for funnel analysis

  • Where does decision-making actually happen?
  • What information do users need during triage?
Funnel chart showing session drop-off between pipeline and submission detail
Funnel analysis across 7.09K sessions.

Questions for heat map

  • Are users navigating efficiently, or compensating for missing context?
  • How can the interface support comparison across submissions?
Annotated analysis of which areas of the submission screen users engage with
Heat map to understand what contents users care the most.
Insight

What the data revealed

  1. Only 46.1% of users entered the submission detail page
  2. The Back button was the most-used interaction on submission detail
  3. Users frequently looped between pipeline and submission detail
  4. Submission detail was used selectively to access specific context, especially summaries and notes

Why Back button dominance?

Fast comprehension first, depth on demand.

The pipeline view functioned as the true decision workspace, while the submission detail page acted as a short-term inspection layer.

Underwriters relied on the pipeline page for orientation and comparison, opening submission detail primarily to verify context before returning to continue their review.

Design

Design direction

  1. 01Rebalanced hierarchy to surface key signals closer to the pipeline.
  2. 02Cut navigation overhead with pre-filtered views and filters-on-demand.
  3. 03A contextual Details drawer for files, AI insights, history and notes.
  4. 04An adaptive viewport for dense analysis without losing context.
Set of redesigned Launchpad screens showing the new navigation model

Reducing noise, preserving control

I designed a 'filters-on-demand' system that reduced visual complexity while maintaining flexibility. Frequently used filters were prioritised, and reusable filter presets allowed underwriters to quickly re-establish context across sessions.

I then consolidated key submission information into a contextual drawer, reachable without leaving the pipeline.

Filters-on-demand: surfacing only what's needed, when it's needed.
Details drawer showing Notes, Activities, Files, Synthesis and Contract tabs beside the pipeline
The Details drawer: files, synthesis, activities and notes in one place.
Multi-select: acting on several submissions in a few clicks.

Adaptive viewport for high-density data

To accommodate dense tables, I designed an adaptive viewport that supports uninterrupted analysis while preserving quick access to workflow context and collaboration tools.

The adaptive detail view, expanding to suit the task in hand.
Dynamic viewport: the workspace adapting as the underwriter moves between comparison and detail.
Impact

Impact

With over 224 hours of monthly usage, even small improvements could scale quickly. Reducing just one minute per session could save over 45 hours per month across underwriting teams.

Due to roadmap priorities, this work was positioned as a future optimisation, establishing a clear, data-backed direction for improving underwriting efficiency at scale.