Casey Lin

work

résumé

Casey Lin

work

résumé

Casey Lin

work

résumé

← All work

Revamping a home insurer's website that had to look bold and read trustworthy.

Revamping a home insurer's website that had to look bold and read trustworthy.

A fast-growing home insurer asked for a website as bold as their racing sponsorship. Home insurance buyers want calm and credibility. I designed for both across a 40+ page rebuild, from the winning proposal to launch.

Role

Sole Product Designer

Timeline

Jan – Aug 2025

Team

PM, account owner, content writer, 2 designers, 2 developers

Type

Marketing Site · AI POC

Outcome

Usability and accessibility has increased largely while maintaining the brand needs for this company

64 → 88

of performance score gain

100 / 100

accessibility score increase from 88 to 100

background

One site, two audiences, one question each.

Users didn’t know where to start. Buyers didn’t know why it mattered.

A homeowner buys insurance hoping to never use it. What they want is proof the company stays when the storm hits. An independent agent asks a different question: is this carrier worth partnering with. One site had to convince both.

A homeowner buys insurance hoping to never use it. What they want is proof the company stays when the storm hits. An independent agent asks a different question: is this carrier worth partnering with. One site had to convince both.

Original Flow before start chatting

We entered a competitive bid and built the proposal for a full website redesign. The team won a $1M+ engagement with a national home insurer serving 250K+ customers at the time.

Over 8 months we rebuilt the site from WordPress to a custom Drupal build, working with a content writer, two developers, and the client's marketing team.

We entered a competitive bid and built the proposal for a full website redesign. The team won a $1M+ engagement with a national home insurer serving 250K+ customers at the time.

Over 8 months we rebuilt the site from WordPress to a custom Drupal build, working with a content writer, two developers, and the client's marketing team.

The old site answered neither audience well. An agent opening the state map found no state names. Clicking their state landed them on a quote form built for homeowners.

The old site answered neither audience well. An agent opening the state map found no state names. Clicking their state landed them on a quote form built for homeowners.

BEFORE THE ENGAGEMENT · 2024 PITCH

I had no prior experience in OCM, so I started by learning the domain before designing anything.

I had no prior experience in OCM, so I started by learning the domain before designing anything.

Before the engagement began, our 2024 pitch had to answer the brief's call for innovation. I designed Orion Assist, a chatbot proof of concept for the quote journey.

Quick replies constrained input where mock data could not carry open conversation. Lead details arrived in steps: name and state first, phone number only mid quote. The idea of uploading a current policy for a faster quote came from my manager, building on our document extraction work. I designed the flow around it: after extraction, a review screen let people correct every field before any quote was made. I also proposed an AI site search to match where the market was heading.

Before the engagement began, our 2024 pitch had to answer the brief's call for innovation. I designed Orion Assist, a chatbot proof of concept for the quote journey.

Quick replies constrained input where mock data could not carry open conversation. Lead details arrived in steps: name and state first, phone number only mid quote. The idea of uploading a current policy for a faster quote came from my manager, building on our document extraction work. I designed the flow around it: after extraction, a review screen let people correct every field before any quote was made. I also proposed an AI site search to match where the market was heading.

Prototype demoed in the 2024 pitch. The build was scoped for a later phase.

The Orion Assist POC drew strong interest from client leadership and contributed to winning the $1M+ engagement. Sales and account leads carried the pitch itself.

The engagement scoped the AI features for a later phase. What follows is what we shipped.

The Orion Assist POC drew strong interest from client leadership and contributed to winning the $1M+ engagement. Sales and account leads carried the pitch itself.

The engagement scoped the AI features for a later phase. What follows is what we shipped.

CONSTRAINTS

What could not move.

Users didn’t know where to start. Buyers didn’t know why it mattered.

PALETTE

Red, black, turquoise, and tan were fixed. 

The turquoise stayed: the client saw it as what set them apart from other red and black brands.

Original landing page

IMAGERY

F1 racing car and galaxy theme imagery is mandatory

F1 sponsorship photos and galaxy imagery were leadership requirements, tied to the company's name.

Original chat entry experience

RESEARCH

The audit found a missing system, and the data reframed the brief.

The audit found a missing system, and the data reframed the brief.

UX Audit, Usability, and Accessibility Review

UX Audit, Usability, and Accessibility Review

We started with a heuristic review of the current site and spotted 56 issues across 27 pages. I owned the accessibility and usability side of the review. Seven button styles, drifting card heights, typography that changed page by page: the site had never had a system to hold it together.

We started with a heuristic review of the current site and spotted 56 issues across 27 pages. I owned the accessibility and usability side of the review. Seven button styles, drifting card heights, typography that changed page by page: the site had never had a system to hold it together.

Two findings set up the decisions that follow. The interactive map showed no state names, and clicking a state jumped to a quote form instead of state information. And the site offered no assistant of any kind for people who could not find what they needed.

Two findings set up the decisions that follow. The interactive map showed no state names, and clicking a state jumped to a quote form instead of state information. And the site offered no assistant of any kind for people who could not find what they needed.

I built two personas from the client's demographic data and desk research. Marcus, an agent, cannot easily judge which carriers are worth partnering with. Sarah, a homeowner in a hurricane state, fears the cancellation letter more than the premium.

I built two personas from the client's demographic data and desk research. Marcus, an agent, cannot easily judge which carriers are worth partnering with. Sarah, a homeowner in a hurricane state, fears the cancellation letter more than the premium.

27-page evaluation report, 56 issues.

27-page evaluation report, 56 issues.

27-page evaluation report, 56 issues.

Income, home value, and geography from the client's customer data.

Over half of customers earned $100K+ and owned higher-value homes. 69% lived in three hurricane-prone states where insurers routinely cancel and withdraw. These buyers were not shopping on price. They were asking whether this company would still be there after the storm. This made us question whether bold and striking was the right brief at all.

Over half of customers earned $100K+ and owned higher-value homes. 69% lived in three hurricane-prone states where insurers routinely cancel and withdraw. These buyers were not shopping on price. They were asking whether this company would still be there after the storm. This made us question whether bold and striking was the right brief at all.

Stylistic Direction

Stylistic Direction

We explored four stylistic directions and converged on two concepts to build out. The other designer built Friendly and Approachable. I designed Elegant and Sleek: rich blacks, restrained contrast, editorial layouts, premium lifestyle imagery. The client's marketing team approved it as the simple, sophisticated direction they had asked for.

When document generation lived inside the chat interface, it increased prompt complexity, strained token limits, and produced inconsistent results. Users couldn’t rely on the output. Engineers also confirmed the architecture couldn’t support both reliably in the same context window.

We evaluated three sitemap structures as a team and shipped a combined architecture with homeowner-first weighting.

Marketing asked for simple yet sophisticated. The CEO asked for bold and energetic. Same company, two briefs.

“OCM people don’t love living in Excel and doing project planning. If we could create our plans really early and quickly, then we can get to the good stuff.” - Cindy, OCM Lead

When we presented to the CEO, his reaction reframed the brief: not bold enough, not energetic enough.

When document generation lived inside the chat interface, it increased prompt complexity, strained token limits, and produced inconsistent results. Users couldn’t rely on the output. Engineers also confirmed the architecture couldn’t support both reliably in the same context window.

Key decisions

Key Decisions

Key Decisions

Decision #1

Give agents their answer in one glance.

Problem: I designed the agent view. Before the redesign, agents could not easily judge whether the company was worth partnering with, because the information they needed was buried across the site. The audit made it concrete: the map had no state labels, and clicking a state sent agents to a homeowner quote form.

Options

The information architecture had one job: answer the disqualifying question first. Can I even sell this in my state. We considered a plain list of states, cheaper to build and easy to scan alphabetically. I rejected it as the lead element because agents think in territories, and a labeled map answers the coverage question in one glance. The audit had also shown exactly how the old map failed: hover-only state names and a click that jumped to a quote form. So the map stayed, rebuilt on its own terms: labels always visible, a custom mobile view instead of hover, and a click that leads to state detail.

Solution: I led with three partnership benefits, followed by an interactive US map showing which states were in the service area, so agents outside those states would not waste time on a partnership they could not use. Clicking a state revealed the products available there and the name of the Territory Manager, so an agent knew both what they could sell and who to call.

Solution: I led with three partnership benefits, followed by an interactive US map showing which states were in the service area, so agents outside those states would not waste time on a partnership they could not use. Clicking a state revealed the products available there and the name of the Territory Manager, so an agent knew both what they could sell and who to call.

Interactive Journey & Task Explorer on the product marketing landing page

impact

What we can measure.

What we can measure.

Measured with Chrome Lighthouse at launch in 2025, desktop performance rose from 64 to 88, a 37% gain, and the accessibility score rose from 88 to 100. The score still holds at 100 as of mid 2026.

The client has since grown past 600K customers, and expanded from 9 to 14 states on the templates we delivered, without a redesign. After handoff, the client launched a new product line on the same page templates.

The agent map outgrew its original job. The client made it the main feature of the public All States page, and it is now how the company presents its coverage to everyone, homeowners included.

What I would measure next: how many agents move from a state page to Partner With Us.

looking back

One lesson I kept.

One lesson I kept.

Lesson 1

Technical constraints shaped the design more than preferences did.

I assumed solid research would change the client's mind. Thirteen interviews said the galaxy imagery confused people, and the client kept it anyway. What changed the outcome was a containment plan that let the constraint stay. Now when persuasion fails, I propose boundaries and keep the work moving.

Lesson 2

Domain experts need something concrete to react to.

—-

Next

Designing a Governed, Recoverable Reassignment

Designing a Governed, Recoverable Reassignment

© 2026. All rights reserved.

Made by

Casey Lin

© 2026. All rights reserved.

Made by

Casey Lin

© 2026. All rights reserved.

Made by

Casey Lin