Roblox · Case study

Roblox · Case study

Universal Search

Universal Search

Universal Search

Unifying search experiences and design language across categories, surfaces and devices.

TIMELINE

2024–2026

ROLE

Lead Designer

DELIVERABLE

Design system

TEAM

Design Operations Designers across four lanes Design Systems Product Engineering Research Design Leadership

Search & discovery

Design systems

Information architecture

Stakeholder alignment

Cross-platform

01

/

12

Universal Search · 2024–2026

One search system for every surface.

02

/

12

The problem

Search worked differently on every surface.

Color, icons, wording and placement all drifted from page to page, and each page only searched its own category. Finding a game, a person and an avatar item meant three separate searches.

A wall of Roblox search screens from phone, desktop, console and VR, each with its own layout, colors and labels.

The audit: search on mobile, desktop, console and VR, captured as it was.

03

/

12

Role and approach

I led search across teams I didn’t manage.

Role

Lead designer for search, partly embedded with the design-systems team. Owned the cross-category audit and the shared pattern.

Partners

Five teams across all four lanes and design systems. Twice-weekly working sessions, a weekly team review and a weekly VP review.

Goal

Define horizontally aligned patterns and UI elements across a diverse set of teams, each maintaining its own.

Audit every surface

Share one vocabulary

Limit to a core set

Build it into the design system

04

/

12

The framework

Three kinds of search cover every search.

Four search fields that search games, people and items together.

Omni: discovery across the product.

Omni: discovery across the product.

Search fields limited to one area.

Scoped: discovery within context.

Scoped: discovery within context.

A search field with filter chips that narrow results.

Filtered: refine known content.

Filtered: refine known content.

05

/

12

One flow

Four states, designed for every surface.

Wireframe phone with the search entry point highlighted.

Pre-query · entry point

Pre-query · entry point

Wireframe phone showing the search landing page.

Pre-query · search landing page

Pre-query · search landing page

Wireframe phone with typeahead suggestions above the keyboard.

Active-query · typeahead

Active-query · typeahead

Wireframe phone showing a results page with tabs.

Post-query · results page

Post-query · results page

06

/

12

Decision one

Anatomy of the search field.

Hint text names the scope before you type.

Secondary actions are optional and sit inside the field.

Entered, typing and cleared look the same on every surface.

One back chevron replaced every back, close and clear button.

Search field in its default state with a back chevron and clear button.

Pre-query: prompt.

Secondary actions: voice, camera, QR code and AI.

Contextual secondary options.

Search field active, then mid-typing.

Active-query: entered, then typing.

Search field with a submitted query.

Post-query: submitted.

07

/

12

Decision two

A small required set keeps the library clean.

Every element was pressure-tested on every surface and device. If it only worked in one place, it was cut, merged or made a variant.

Game, user, item and avatar tiles.

Cross-platform content tiles.

Cross-platform content tiles.

Suggestion rows: recent, autocomplete, game, friend, avatar item and more.

Rich predictive list items.

Rich predictive list items.

Search chips: suggested, popular, event, trending, creator and more.

Dynamic content chips.

Dynamic content chips.

08

/

12

The solution

Results regroup as you type.

Experiences, marketplace items and people share one typeahead pattern. Suggestions regroup as the query changes, so a person can reach any of them from the same field.

Phone typing “race”, suggesting experiences.

Experiences.

Experiences.

Phone typing “Anime”, suggesting marketplace items.

Marketplace.

Marketplace.

Phone typing “gra”, suggesting people.

People.

People.

09

/

12

The solution

The empty state becomes a place to discover.

The landing page replaces a blank screen with recently visited and trending searches, and adapts from phone to desktop. An MVP delivered both.

What people see before they type is its own case study:

Two phones showing the search landing page: recently visited, trending searches, explore and trending items.

Recently visited, trending searches and trending items, before anyone types.

10

/

12

The system

Same pattern, narrower scope.

Discovery searches across games, people and items. The marketplace runs the same flow, scoped to avatar items.

Discovery search: landing, typeahead and results across experiences, avatars and people.

Omni search · discovery.

Marketplace search: the same flow scoped to clothing and accessories.

Scoped search · marketplace.

11

/

12

The outcome

The deliverable was the system: UI, interactions and patterns.

In the design system

The pattern is defined and documented, and the components are built into the design system code, ready for any page to pull in.

Alignment

Five teams brought together around one pattern, then senior leadership sign-off.

Built to extend

Pages can align as they are updated. People search, creator surfaces and a unified backend can build on the same pattern.

Built to extend

Pages can align as they are updated. People search, creator surfaces and a unified backend can build on the same pattern.

The search component set on blue: the search field, two columns of suggestion rows, a group of chips and game, user, item and avatar tiles.

The component set: search field, suggestion and search-term rows, chips and tiles.

12

/

12

Reflection

What I took from it.

Building consensus across orgs with very different cultures was the core of the work.

Scoping around years of technical incompatibility let the system move without waiting on a unified backend.

The model I’d run from day one: drive the system forward with design-org buy-in, and keep the technical roadmap informed but separate.

I built allies by making the case for the system: cleaner code, consistent assets, fewer parts to maintain, faster iteration.

© 2026 Ian Khoogar-Kirk

© 2026 Ian Khoogar-Kirk

© 2026 Ian Khoogar-Kirk