Terminology governance
at scale
When a product ships without shared language, every team invents their own. I built Meta's wrist wearable terminology governance from scratch — 90+ terms across hardware, health, UX, and AI, with rules every writer, engineer, and localization partner could actually use.
Terminology Governance
When a product ships without shared language, every team invents their own. I built the system that stops that — 90+ terms with rules every writer, engineer, and localization partner can use.
No shared language.
Every team inventing its own.
On the wrist wearable, engineering called a UI element one thing, design called it another, and product marketing used a third. Localization teams were translating inconsistent source copy. Health features had terminology that needed FDA alignment. And new terms were being coined in meetings and Figma files faster than anyone could track them.
There was no terminology system. No single source of truth. No process for deciding what we call things.
I was not asked to do this. I saw the problem, scoped the solution, and built it.
Audit everything.
Then build the system.
I started by auditing every term in use across the product — hardware names, software labels, health feature terminology, UI component names, settings categories. I found inconsistencies everywhere: the same feature referred to by different names in the app, on the watch, and in marketing materials.
I met with content designers across product lines to combine and audit learnings and find patterns. I worked with localization managers to understand what was breaking in translation. I partnered with Legal and product marketing to understand which names had trademark constraints.
Then I built the governance system: defined terms with approved usage and deprecated alternatives, created concrete "use this, not that" rules (not style-guide abstractions), aligned health terminology with regulatory requirements, and converged competing systems into single umbrella terms. I also built the governance process itself — how new terms get proposed, reviewed, and approved, so the system maintains itself without me.
A system that
runs without me.
Approved usage, deprecated alternatives, and context for when each applies — covering hardware, health, UX, and AI terminology across the product.
A process for proposing, reviewing, and approving new terms — so the system maintains itself and doesn't require a human lookup table.
Guidance tested across Danish, Spanish (Spain), Spanish (Latin America), French, Italian, Japanese, and Korean — reducing translation rework and preventing drift across markets.
One spreadsheet.
Five domains. 90+ terms.
The global glossary wasn't a document — it was a system. Five sheets, each covering a distinct domain, each with the same structure: term, description, whether it appears in UI, whether it's in Acrolinx (Meta's terminology enforcement tool), usage guidance, localization flag, owner, and status. Every term traceable. Every decision documented.
Each term also tracked its Acrolinx status — Meta's terminology enforcement tool that flags non-preferred terms in real-time as writers type. Getting a term into Acrolinx isn't just documentation. It's infrastructure: the term becomes enforced, not just recommended.
Every decision documented.
With the reasoning.
Terminology governance isn't just about picking the right word. It's about documenting why — so the next person doesn't have to relitigate the same decision. The content decisions log tracked every call, with context, rationale, and who owned it. A few representative examples:
✗ Don't use for system states the user didn't trigger: "Incoming audio call" (no ellipsis — it's declarative, not in-progress)
✗ Avoid "powered on," "switched on," "enabled" (unless in a settings toggle context)
✓ Use "Messages" as the label for the native SMS-only app
✗ Don't use "Messaging" to refer to the SMS app specifically
✗ Don't use "TalkBack" (Android-specific brand name)
✓ Do include AM/PM in notifications, reminders, and time-specific contexts
✓ Use "Payments" in India only (regulatory requirement)
Not just building it.
Using it.
Meta's wearables org has a dedicated terminology governance function — a terms lead who maintains the org-wide glossary and runs weekly office hours. I wasn't just building my own system in isolation. I was a regular participant in that governance process, bringing wrist wearable terminology questions for alignment and contributing to the shared infrastructure.
Three sessions from the terms lead's own notes:
Late naming changes
virtually eliminated.
Three competing terminology systems converged into one on the wrist wearable — for example, "widget" became the umbrella term, deprecating "complication" and "glanceable." Forty-four settings UI copy were standardized across app and hardware — the full language standard (including the four rules and decision rationale) lives in the Settings IA case study. Late naming changes during content lock were virtually eliminated.
Localization rework dropped — stable source copy meant fewer translation cycles. The system was adopted by adjacent teams and extended to a new AI product line.
The lesson I carried forward: if people keep coming to you with the same question, the answer isn't to keep answering — it's to build the thing that answers for you.
"Ashlee has been our steward for terminology, including preventing the team from naming regressions introduced by mistake and/or due to the product process being so lengthy."
— Product design peer"Thanks for spearheading this work and getting the team to align and document terminology. This is going to be so valuable for our team as we grow and look to stay aligned across the product."
— Content design peer