New users
Orientation, guidance, simplified discovery, and clearer first paths into the ecosystem.
Globant, big tech client
Enterprise information architecture
Enterprise navigation and taxonomy for a 70,000 page ecosystem
How I turned fragmented ownership and overlapping product journeys into a navigation system the organisation could govern itself.
A global technology company had grown to more than 70,000 live pages across its products, departments and campaigns. No two teams gave the same answer about where a thing belonged, and customers paid for that disagreement every time they went looking for something.
The menu was the smallest part of the job. Departments with competing incentives had to settle on a handful of shared rules first, and AI read the content at a scale no manual audit could reach, so the remaining arguments were about judgement instead of inventory. What the company kept afterwards was a structure its own teams could run without me.
70,000+
live pages in one ecosystem
1 product
filed under three different names
5 rules
left behind to keep it clean
"Transform structured chaos into something that can be easily managed by an organization."
Where it started
Seventy thousand pages, and no two teams agreed where anything belonged
Every product launch and campaign had added pages, and each one had been filed by whoever happened to own it that quarter. The same product appeared under three names. Related services sat in unrelated places. Customers could not guess where anything lived because the structure recorded the company's history rather than their needs.
Reorganising that by hand would have taken a year and been stale on delivery. We needed to see the whole estate at once, get the owners to agree on a few rules, and leave behind something that stopped the sprawl returning.
What had to be true
Five rules the structure had to keep as the company grew
- Unify the ecosystem: reduce disconnected paths and make product relationships easier to understand.
- Improve discoverability: help users orient quickly and reach relevant content with fewer dead ends.
- Clarify terminology: move away from internal language where it conflicted with user expectations.
- Support governance: give teams principles and decision rules for future navigation changes.
- Prepare for personalization: define experience layers that could evolve with user maturity and context.
Getting people to agree
The structure could not settle until the teams did
Every department wanted its own work one click from the homepage, and each had a fair reason. A recommendation that ignored those incentives would have been overridden within a quarter, so the structure had to give teams a way to settle the next disagreement themselves.
In workshops we made the trade-offs explicit and wrote them down as navigation principles with named owners and a measure attached. Once a rule existed, arguments moved from taste to evidence, and teams could apply it without calling a meeting.
Method
Reading 70,000 pages without reading 70,000 pages
Card sorting works for a few hundred pages. At 70,000 it tells you nothing. Turning every page into a semantic embedding let us group content by what it actually said, which surfaced duplicate coverage, abandoned pages and natural clusters the org chart had hidden.
The clustering made the scale legible. It did not decide anything: every final grouping was checked against business context and signed off by the people who owned the content.
Going deeper
Navigation that meets people where they already are
A first-time visitor and a returning account holder need different things from the same menu. We defined three levels of familiarity so the navigation could adapt later without rebuilding the taxonomy underneath it.
Contextual users
Device-aware navigation, localization, accessibility adjustments, and contextual recommendations.
Known users
Personalized pathways, account-aware navigation, dynamic recommendations, and tailored calls to action.
My role
Turning a standing argument into something people could point at
I owned the IA strategy and the taxonomy, ran the clustering workflow, and facilitated the sessions where department owners, marketing and UX had to reach a decision together.
Most of the value came from making the abstract visible. Once a taxonomy model was on screen, stakeholders stopped debating principles and started pointing at specific branches, which is when the decisions finally moved.
What made it hard
- Organizational complexity: departments had competing expectations around ownership and visibility.
- Scale: the content volume made manual categorization impractical.
- Terminology: product and platform labels were inconsistent across journeys.
- Sustainability: the taxonomy needed to remain useful after the project ended.
What it produced
A structure that outlives the people who built it
The client got a taxonomy that holds at 70,000 pages, far less duplicate and dead content, and navigation paths that follow how people actually search. It maps onto their CMS, so filing a new page is now an obvious act rather than a judgement call.
The structure will drift again as any live system does. The difference is that the company now has the rules and the owners to catch it, which is the part that keeps paying off.
Looking back
Enterprise UX at this scale is mostly organisational work wearing a design brief. The hardest hour of the project was not the taxonomy. It was the workshop where two directors discovered they had been using the same word to mean different things for three years.
This project was completed through Globant and GUT for a large enterprise technology client. Due to confidentiality agreements, specific company details, metrics, and proprietary assets have been omitted or anonymized.