
Content Ecosystem (CES) — Case Study Narrative
Overview
The Content Ecosystem (CES) is a large-scale, headless content repository of over 10,000 articles, videos, and images sourced from internal authors and licensed partners. Every asset entering CES needed to be tagged accurately and consistently using a unified taxonomy to support search, personalization, and downstream content delivery across multiple enterprise portals.
As a taxonomist embedded in a scrum team — and drawing on my background in UX and library science — I governed, tuned, and operationalized the Unified Taxonomy that powered CES. My work spanned taxonomy design, metadata governance, model tuning, automation, reporting, and UX improvements to internal tools.
Over two years, I took ownership of a strong but undocumented taxonomy that had been dormant for nearly a year, developing the automation, tuning workflows, and governance structures that allowed it to operate reliably at enterprise scale.
Context & Problem Space
Before this work began, CES relied on a Unified Taxonomy comprised of seven vocabularies created by a previous taxonomist. The Topics taxonomy was triggered through an AEM workflow, with the goal of automatically applying Topic tags at content ingestion. The remaining vocabularies were applied through a structured Excel export-and-reimport process managed by the technical product manager, referred to internally as the “Bulk Tool,” which compensated for AEM’s lack of a native bulk update feature.
While workable at smaller scale, this hybrid approach became increasingly difficult to sustain as the content library grew.
The challenges were significant:
- Manual tagging was slow and error-prone; tagging 300 assets took approximately one month.
- Metadata inconsistencies led to mistagged, over-tagged, or untagged content.
- The taxonomy lacked documented governance, allowing drift and duplication over time.
- Internal tools were built for developers rather than authors, creating UX and accessibility issues.
- Downstream systems depended on accurate metadata, increasing operational risk.
- The business required a taxonomy model that could scale, automate tagging, and support governance without disrupting existing workflows.
My Role
I served as both taxonomist and UX designer, responsible for:
- Leading ongoing development and governance of the Unified Taxonomy in Semaphore, with primary responsibility for Unified Taxonomy – Topic
- Designing and tuning the topic model that powered automated tagging
- Creating and governing metadata standards for 10,000+ assets
- Building reporting workflows to identify mistagged or over-tagged content
- Automating a 35-step Excel routine using CoPilot Agents to accelerate reporting
- Designing UX improvements for internal authoring tools and workflows
- Creating the governance process for taxonomy intake, evaluation, approval, and deployment
- Training authors and strategists on metadata best practices and tooling
This hybrid role allowed me to bridge UX, taxonomy, and data — designing systems that were both technically sound and human-centered.
Challenges
Several constraints shaped the work:
- Scaling governance of an existing Unified Taxonomy within a growing enterprise content ecosystem
- Navigating hybrid metadata workflows: workflow-triggered Topics tagging and structured Excel-based reimport processes
- Tooling constraints within AEM requiring coordinated workarounds
- Governing and tuning the Unified Taxonomy in Semaphore
- Developing hands-on expertise in Semaphore through self-directed learning while supporting live workflows
- High-stakes metadata accuracy, with errors propagating across multiple enterprise portals
- UX improvements constrained by technical and taxonomy implementation sequencing
These challenges required systems thinking, technical fluency, and a deep understanding of both user needs and metadata behavior.
Approach
I approached this work as an ongoing governance and scalability effort within an existing Unified Taxonomy. My primary focus was Unified Taxonomy – Topic, where accuracy and consistency had the greatest downstream impact, and where partial automation was being explored.
Because metadata application relied on a hybrid process—workflow-triggered Topics tagging in AEM and a structured Excel export-and-reimport process (“Bulk Tool”) for other vocabularies—I prioritized repeatable practices that reduced risk, improved consistency, and supported future automation.
1. Assessing and Strengthening the Unified Taxonomy
Rather than building a taxonomy from scratch, I began by analyzing and validating the existing Unified Taxonomy model in Semaphore to understand how it functioned in practice.
- Reviewed how the vocabularies comprising the Unified Taxonomy operated together within the Unified Taxonomy
- Mapped term relationships and overlaps across vocabularies
- Identified inconsistencies and opportunities for refinement
- Modeled refinements and adjustments in Semaphore
- Helped ensure controlled vocabularies (Topics, Audience, Purpose, Language, Content Type) functioned cohesively
This work strengthened the structural integrity of the Unified Taxonomy and created a clearer foundation for consistent tagging and future automation.
2. Manual Tagging → Automated Tagging
Early in the project, I manually tagged 300 assets — a process that took approximately one month. This experience revealed:
- The true scale of the problem
- The need for automation
- The importance of accurate model tuning
Once the topic model was configured, I could trigger the AEM workflow to tag content in minutes, freeing me to focus on higher-value work such as tuning and governance.
3. Model Tuning & XML Optimization
Model tuning became a core part of my role:
- Analyzed tagging accuracy
- Identified un‑tagged, mis‑tagged, and over‑tagged assets
- Adjusted term weights
- Edited XML rules in Semaphore
- Maintained a .75 threshold to control over‑tagging
This iterative tuning improved tagging precision and reduced the need for manual corrections.
4. Reporting & Automation
To monitor tagging quality, I developed a 35-step Excel routine that:
- Identified un‑tagged assets
- Flagged over‑tagged fragments
- Detected syntax errors
- Calculated monthly metadata health metrics
The routine took about 15 minutes per file — until I automated it.
Using CoPilot Agents, I transformed the routine into an automated workflow that completed the same analysis in seconds. This dramatically improved reporting speed and allowed for more frequent monitoring.
5. Governance Process Design
- I created the governance process for the Unified Taxonomy, including:
- Intake of candidate terms
- Tracking and documentation
- Research and validation
- SME card sorts
- Approval workflows
- Updates to Semaphore
- Testing new terms in AEM
- Weight calibration for automated tagging
This governance model ensured long-term consistency and controlled vocabulary growth.
6. UX Improvements to Internal Tools
Because the tools used to manage metadata were originally built for developers, I provided UX guidance to improve:
- Field labeling and grouping
- Accessibility
- Navigation
- Error handling
- Author/developer mode separation (removing “easter egg” interactions)
These improvements reduced friction for authors and made the system more usable.
Outcomes & Impact
Operational Impact
- Reduced tagging time for large content sets from approximately one month of manual work to minutes via workflow-triggered automation
- Improved metadata consistency across 10,000+ assets
- Reduced over-tagging, syntax errors, and manual correction cycles through model tuning
- Established repeatable reporting processes, enabling proactive detection of tagging issues
- Improved downstream content delivery accuracy across enterprise portals
Team Impact
- Developed job aids and documentation supporting authors working in AEM and, when appropriate, the structured Excel reimport process (“Bulk Tool”)
- Conducted frequent in-person training sessions, increasing author confidence in applying and managing metadata
- Served as a bridge between technical and non-technical teams, translating taxonomy and tooling concepts into accessible guidance
- Provided documentation leveraged by engineering leadership for cross-team communication
- Supplied developers with clear, actionable requirements for tool and workflow improvements
Organizational Impact
- Strengthened the Unified Taxonomy as a foundational enterprise system asset
- Reduced operational overhead through workflow automation and improved governance
- Established a governance model supporting sustainable vocabulary growth
- Improved metadata quality supporting search, personalization, and content reuse initiatives
- Partnered with content strategy to develop a plan for broader enterprise adoption of AEM and Semaphore
Reflection
The Content Ecosystem project marked a pivotal moment in my career — where my UX background and library science training converged into a hybrid practice of taxonomy, metadata, and systems design.
It reinforced several strengths:
- Designing clarity within complex content systems
- Bridging UX and information science
- Creating scalable metadata models
- Automating workflows to reduce operational burden
- Leading governance for long-term sustainability
- Improving internal tools for non-technical users
It also reinforced the importance of sequencing UX work early — and navigating situations where technical implementation precedes design.
Ultimately, this project solidified my identity as a systems thinker who thrives at the intersection of UX, taxonomy, and data.