Get Work
Redesigned a national work queue used by 450k+ people. Researched, strategized, shipped.
One ticket. Three years of the same story.
One user flagged a filter issue during an open-hours help session. While looking into the ServiceNow data, what we found was bigger than one broken filter.
Roughly 12% of tasks pulled through the system were unworkable. Users had to return them to the queue, creating duplicate work and eroding trust in the platform. Three years of support tickets told the same story in different words.
Three root causes happening at once
-
01
Excluded user groups. The filtering system was built for a narrow set of workflows. Users with different roles had no path through it.
-
02
Unreliable data at the source. Labeling was inconsistent at the backend level. Filtered results couldn't be trusted, so users stopped trusting them.
-
03
No room to scale. The architecture couldn't absorb growth. Every new form type made the problem worse.
What I owned
Complex problems need great teams.-
Led the end-to-end redesign as UX Designer on a cross-functional team of 17
-
Analyzed ServiceNow tickets that surfaced the 12% unworkable task rate
-
Conducted contextual inquiry sessions and SME interviews across product lines
-
Audited the information architecture across all form types and flagged critical labeling inconsistencies
-
Designed the 3-tier filter system and contributed the chip filter as a new design system component
-
Set up behavioral tracking in New Relic post-launch and ran pulse surveys to measure impact
What I owned
Complex problems need great teams.-
Led the end-to-end redesign as UX Designer on a cross-functional team of 17
-
Analyzed ServiceNow tickets that surfaced the 12% unworkable task rate
-
Conducted contextual inquiry sessions and SME interviews across product lines
-
Audited the information architecture across all form types and flagged critical labeling inconsistencies
-
Designed the 3-tier filter system and contributed the chip filter as a new design system component
-
Set up behavioral tracking in New Relic post-launch and ran pulse surveys to measure impact
We dug deeper than the original ticket.
From rigid cascading dropdowns to a dynamic, role-aware filtering system
We started with the data. Manually combing through ServiceNow ticket queries was not glamorous work. But it gave us something more valuable than assumptions. It gave us patterns.
Users weren't just frustrated with one filter. They had built workarounds around the entire system. Some were cross-referencing dashboards just to verify their own results. Others were navigating five dependent dropdowns before finding out there was nothing in the queue for them.
"This filter doesn't fit my work. I have to go through all the dropdowns just to scroll to the bottom and find out there's nothing there for me."
Contextual inquiry participant
"I always cross-reference the dashboard to make sure the numbers are actually right."
Contextual inquiry participant
Contextual inquiry sessions made the frustrations and workarounds visible, while SME interviews across product lines revealed something the original design had missed entirely. Responsibilities and workflows varied significantly across roles. The system had assumed uniform processes, excluding entire groups of users with different workflows.
During a solo IA audit, I noticed the filter labels were inconsistent across form types. Same concept, different names depending on where you looked. I documented it with screenshots and brought it to the full team. The PM was aware of some inconsistencies but hadn't mapped the full scope. Once the dev team saw it laid out visually, it unlocked a round of clarifying questions that tightened the entire information architecture. One observation became a structural fix.
After the initial designs took shape, we mapped the IA across all form types. This gave the team a shared view of where filters overlapped, conflicted, or were missing entirely, and helped devs visualize the filtering relationships before build.
Three calls that shaped the outcome.
Before touching the interface, we made two strategic calls. Fix the data foundation first. Roll out in phases to protect a mission-critical system. Everything else followed from those two decisions.
A 3-tier filter architecture over a flat redesign
The temptation was to clean up the existing dropdowns. We didn't. The problem wasn't the visual design. It was the underlying structure. We replaced cascading dropdowns with a 3-tier filter panel built around how users actually work.
-
Tier 1
Location and required form or task type. Auto-populated. Always visible.
-
Tier 2
Expanded filter options that appear based on Tier 1 selections.
-
Tier 3
Additional queue-specific filters that surface only when relevant.
The structure was driven by both user needs and backend performance, keeping data loading fast across 100+ form types.
The 3-tier structure supported every workflow type without sacrificing backend performance
Chip filtering over toggle switches
We explored toggle switches early. They looked clean. But toggles imply instant feedback and our filtering required a deliberate apply action. They weren't the right pattern.
Chip filtering let users see all available options at once without navigating dropdowns. This component, applied where 2 to 3 options exist, reduced clicks and improved scannability. Since it didn't exist in the design system, a fully documented and reusable component was added to the library. It's now available across the platform beyond Get Work.
Component anatomy and specs documented for the design system
Default, hover, and selected states
Fixing the labels before fixing the interface
This was the call that mattered most. We could have shipped a cleaner UI on top of inconsistent data and it would have failed. Standardizing the backend labeling before launch meant the filtering results were trustworthy from day one.
Trust in the data was what users had lost. Getting it back required fixing the source, not just the surface.
The response was measurable.
unworkable task rate addressed
saved per filter interaction
daily users on the platform
Two previously unsupported user groups gained reliable access to their work for the first time. They stopped building workarounds. Support tickets decreased.
Users with previously unsupported workflows no longer needed workarounds and felt included for the first time.
Post-launch pulse survey
The chip filter component is now part of the platform library. Other teams can use it. The work outlasted the project.
The 3-tier structure can absorb new form types and workflows without a structural overhaul. Built to grow.
A supervisor distribution feature would have let managers assign work directly from the queue. It was scoped out to keep the project focused. Sometimes the best design decision is knowing what not to build yet. It's documented. It's on the roadmap.
What stayed with me
-
One support ticket opened a three-year problem. The best research doesn't always start with a research plan.
-
Fixing the data before fixing the interface wasn't the obvious call. It was the right one. Surface-level solutions on broken foundations don't ship, they just delay the failure.
-
The best moment on this project wasn't a design decision. It was a room full of designers, PMs, and devs looking at the same screenshots and finally seeing the same problem.