Information Architecture
Country selector content strategy
Google Operations Center · Oct 2024 – Present
250+ selectors · 99.64% quality score · 300+ assets governed in GitLab.
Designed the user research, set the portfolio-level content strategy, and authored and governed 300+ assets in GitLab.
HTML · CSS · GitLab · Markdown · XML · MadCap Flare
- Problem
- 250+ country selectors, each treated as its own content problem, with no portfolio-level view of what differed and why.
- Insight
- Every proposed improvement was being argued from one market's evidence, because no one had ever looked at the selectors as a single portfolio.
- Result
- Targeted user research and a portfolio-level content strategy spanning every selector, held to a 99.64% quality score.
The full story
Google's billing and payments surface reaches advertisers through 250+ country selectors, each carrying its own payment methods, regulatory constraints, and reader expectations. The content had grown the way the markets had: one selector at a time. There was no portfolio-level view of what genuinely differed between them and what merely looked different.
The work started with research, not writing. Designing targeted user studies across the selector portfolio produced something the per-selector view could not: evidence about where readers actually failed, and which differences between markets were substantive and which were leftovers of history. The content strategy was built on that evidence, market by market.
Execution ran through 300+ documentation assets built in HTML and CSS and governed through GitLab version control, which put documentation changes under the same review, history, and rollback that engineering already had. The quality score across that output is 99.64%, sustained at the publishing velocity the surface requires.
The number that is easy to misread is the 99.64%. It is not a measure of careful proofreading. At this volume it is a measure of whether the system catches defects, because no individual writer's attention scales to 250 markets and 300 assets.
Key decisions
- Do user research before content strategy. A strategy that is not grounded in where readers actually fail is a reorganization, and reorganizations cost more than they return.
- Govern the assets in GitLab instead of a CMS-native workflow. Version control gives documentation the review history and rollback that the publishing cadence requires, and it makes a defect traceable to the change that introduced it.
- Treat portfolio-level consistency as the unit of quality, ahead of per-article polish. At 250+ selectors, a reader's experience is shaped by whether the markets agree with each other, not by whether any single page is well turned.
What it taught me
The 99.64% is a review score. It says the content met the standard it was measured against, which is worth having, but it does not tell me whether an advertiser in a market I have never been to understood their invoice. Closing that gap is the part of this job I still have not found a good way to instrument.