Unified Expression System
3 billion people. 6 reactions. All of them broken in different ways across different surfaces.
Every surface at Meta — Facebook feed, Instagram Stories, Messenger, Reels, Comments — had evolved its own reaction system independently. Six different implementations. One company. No shared foundation. That's not a design problem. That's a design debt emergency at 3 billion user scale.
Setting the Scene
Expressions and reactions are how billions of people communicate emotion online. A thumbs-up. A heart. A crying-laughing face. These are not trivial UX details — they're the primary vocabulary of human connection at internet scale. And at Meta, that vocabulary had fragmented. Each surface team had made their own decisions, leading to inconsistencies that users felt even when they couldn't name them.
The Problem
The fragmentation meant: users got different experiences depending on which surface they were on. New surfaces couldn't launch with a coherent expression system because there wasn't one. Accessibility was an afterthought — the emoji picker was keyboard-unfriendly and screen readers struggled throughout. And cultural nuance was almost entirely absent — one set of reactions, 130+ countries, wildly different meanings.
What made it hard
- →Any change to reactions affected billions of people — migration complexity was enormous
- →The legacy system had 6 core reactions with different implementations across every surface
- →Localization requirements: 130+ languages and regions, each with culturally specific expression norms
- →Performance non-negotiable: expression UI had to load in under 300ms even on slow networks
- →Accessibility was broken and needed to be retrofitted, not designed in
- →Competing stakeholder priorities across multiple surface teams with no shared roadmap
The biggest shift in my thinking: expressions aren't universal — they're deeply cultural. The problem wasn't 'design one system.' It was 'design a system that enables systems.' One-size-fits-all would fail. A flexible framework could scale.
What the Research Revealed
- →The 'Haha' reaction was consistently misinterpreted across markets — users read it as mockery, not laughter
- →45% of users in certain Asian markets strongly preferred custom expressions over the preset global reactions
- →The emoji picker was keyboard-unfriendly — keyboard-only users had almost no way to navigate it efficiently
- →Users wanted multi-reaction capability (expressing 'Wow' and 'Love' simultaneously) — UI limited them to one
- →Thumbs-up worked universally; several gesture-based emojis were offensive in specific regions
- →200+ interviews across 10 countries surfaced starkly different emotional communication norms
The Approach
I designed a three-tier expression framework: Universal Core (7 expressions that resonate across all cultures), Regional Variants (culture-specific expressions controlled by regional teams), and Extended Reactions (user-generated custom reactions and emoji pickers). The architecture was designed to be scalable, governable, and adaptable — and critically, to prevent the same fragmentation from happening again.
Three tiers instead of one-size-fits-all
Universal Core for global consistency, Regional Variants for cultural flexibility, Extended Reactions for user expression. Each tier had its own governance model so regional teams could customize without breaking the system.
Accessibility-first emoji picker redesign
Search-first UX, keyboard shortcuts, proper ARIA labels. Designing for screen readers and keyboard-only users actually improved the experience for everyone — faster, more organized, less hunting.
Multi-reaction support
Users could now apply multiple reactions to a single item instead of replacing the previous one. More expressive. More human. Adopted by 45% of users in the first month alone.
Standardized expression component
One component, all surfaces. Reactions looked and behaved identically across Facebook, Instagram, and Messenger. The foundation that would prevent future fragmentation.
Proving It Out
Regional pilot testing across 100M users in 5 diverse regions over 4 weeks. A/B test of the emoji picker redesign vs. the original at 50M users. Ongoing accessibility testing with blind users, deaf-blind users, and motor-impaired users throughout the process. And 50+ qualitative sessions with diverse international users to capture what the numbers couldn't.
- →Regional variants saw 3.2x higher engagement in target regions vs. universal-only reactions — the flexibility hypothesis proved out
- →Multi-reaction was adopted by 45% of users within the first month, dramatically exceeding expectations
- →Accessibility improvements resulted in a 4x increase in expression usage among screen reader users
- →Emoji picker search reduced time-to-reaction by 60% — people could find what they wanted instead of hunting
Beyond the Numbers
- →Regional teams felt genuinely empowered — they could customize for their markets without waiting for a central design team
- →Users said multi-reaction felt more authentic to how they actually felt about content
- →Accessibility improvements received public recognition and positive press coverage — the work mattered beyond Meta
- →The expression system became the design foundation for reactions in Reels, custom reactions in Stories, and future surfaces
What I Took Away
Systems designed at scale need built-in flexibility. The three-tier framework succeeded because it was built to be adapted — not locked. And accessibility built from day one would have been far cheaper than what we had to retrofit. That's a lesson I won't repeat.
- →Cultural nuance matters more than we typically acknowledge in design — working with regional teams transformed our understanding from 'universal emotions' to 'culturally grounded communication'
- →Designing for accessibility first actually improves UX for everyone — it's not a constraint, it's a design forcing function
- →Engagement metrics can hide important insights — the numbers showed success, but qualitative feedback revealed how much multi-reaction and regional customization actually meant to real people
- →Governance is part of the design. Without a clear model for how regional teams could customize, the system would have fragmented again
What's Coming Next
Custom reaction capability — let users upload their own emojis. Reaction trending to show which reactions are surging in real time. A creator analytics dashboard for communities to see expression patterns. AI-suggested reactions based on content understanding. And extending the framework to every surface that launches at Meta from here on.