UX WORK › CASE STUDY
One Size Does Not Fit All.
Embedded Files
Restructuring Indeed's segmentation platform to unlock 30× experimentation.
Role
UX Lead
Timeline
February 2022 — January 2024
Scope
End-to-end platform redesign, information architecture, UX research, design systems
Team
20+ Segment Managers, 1 Design Technologist, 1 Product Manager, 6 Engineering partners
TL;DR
The problem
The Segmentation platform was the infrastructure behind Indeed's #IndeedGetsMe mission but Segment Managers were spending most of their time on repetitive setup work, duplicating segments, switching between 11+ tools, and managing cross-team handoffs, because of how the system was structured. This pulled them away from their core work.
What I did
Led end-to-end platform redesign, research, IA, Figma specs, and a phased delivery strategy, proposing a fundamental restructure of how segments, levers, and experiments related to each other.
The key insight
The problem was structural. The platform was built by engineers for engineers, but Segment Managers aren’t as technical. Aligning the architecture to their mental model changed everything.
The result
Experimentation grew from 320 to 9,560 experiments per year; a 30× increase. Tools required shrank from ~13 to 3–4. Core workflow inputs cut by 29%.
OUTCOMES
30×
growth in experiments per year
320 → 9,560
↓70%
consolidation of tools required
~13 tools → 3–4
↓29%
fewer inputs in core workflows
7 inputs → 5
THE PROBLEM
The system was fighting its own scalability.
Segment Managers are responsible for tailoring job seeker experiences on Indeed, but the platform’s architecture buried them in repetitive setup work just to get their job done.
The Segmentation team's mission was clear: meet the unique needs of key segments and occupations, so that job seekers (nurses, teachers, warehouse workers, etc) could say #IndeedGetsMe. That's a company-wide outcome. But the platform Segment Managers relied on to get there was fighting them at every step.
The platform tied segments to individual surfaces across Indeed. Each surface had its own configurable settings baked in. Want to test two different filters for nursing job seekers on the search results page? That's two duplicated segments. Run those tests in parallel? More duplication. A single set of experiments could require 11+ tools and multiple handoffs across systems and teams.
A look at the existing experience where even one workflow required moving across multiple tools and windows, highlighting the fragmentation the new system needed to solve.
MY ROLE
Research, architecture, and a roadmap teams could trust.
I led UX from discovery through phased delivery, and designed the strategic framework that made a multi-year transformation feel achievable.
That meant conducting and scripting UX research sessions with Segment Managers, driving alignment on shared terminology across Product and Engineering, creating the information architecture for the new system, and designing Figma specs that doubled as engineering tickets and roadmap markers; collapsing the gap between design handoff and development scoping.
The system architecture that made the transformation possible, aligning segments, experiments, and levers into a single, scalable model teams could build on.
DISCOVERY
An architectural problem nobody named.
Segments, levers, and experiments (components Segment Managers tweaked) were all tangled together, defined relative to surfaces rather than to the segments (group of occupations) they were meant to serve.
This meant the same segment had to exist in multiple copies just to enable parallel testing. Every new experiment required recreating prior work. The duplication wasn't a workflow quirk, it was load-bearing. The architecture required it.
After hours of research with Segment Managers, one thing became clear: they don’t think in terms of surfaces. They think about people and segments: nurses, teachers, warehouse workers. The platform was structured in a way that didn’t match how they think.
THE DESIGN
Decouple.
Reuse.
Scale.
THE ARCHITECTURAL SHIFT
Make the platform surface-agnostic by separating segments, levers, and experiments into distinct, reusable building blocks, with segments at the core.
This direction came directly from the research, it matched exactly how Segment Managers already thought about organizing their work. It was the right call and a hard one. A change of this scope couldn't ship all at once.
A PHASED DELIVERY TO MAKE COMPLEXITY FEEL ACHIEVABLE
Working backwards from the ideal end state, I developed a phased strategy where every phase was a meaningful, usable milestone, not a percentage of completion. We called this the Minimum Viable Experience (MVE).
The metaphor gave the team a shared language for prioritization conversations, and helped a 3-month phased delivery strategy feel intentional rather than arbitrary.
KEY DESIGN DECISIONS
A unified experience where that same workflow is brought into a single, structured system, reducing fragmentation and making it easier to complete tasks end to end.
Redesigned landing page: orientation before action, surfacing context before asking for input.
RESULTS
The foundation for personalization at Indeed.
The platform became a durable, scalable foundation for personalization at Indeed and cleared the path for Segment Managers to actually pursue the #IndeedGetsMe mission rather than fight their own tools to get there. The work that Segment Managers were uniquely qualified to do was no longer blocked by configuration overhead the system had been generating.
REFLECTION
Following company-wide reorganizations, I transitioned to a new team before the remaining phases could be completed. The phased rollout strategy, documentation, and Figma infrastructure I left behind were designed specifically to give future teams a clear path forward, and the foundation held.
TAKEAWAY
The thing I'm most proud of isn't a metric.
It's that the team understood where they were going at every step, even when the destination was two or three phases away. Our nuanced way of working created clarity that kept the work from drifting. Or stopping.
Google Sites
Report abuse