A content-portfolio framework for a world where documentation has two audiences at once: people and AI agents.
- My role: Creator
- What it is: A content intelligence framework and v1.0 specification
- Built for: Documentation teams deciding where to invest
The problem
Content teams judge success by page views, time on page, and search traffic. But documentation is increasingly read by copilots, retrieval systems, and agents, and those readers don’t show up in traditional dashboards. That creates three risks:
- AI-critical content looks unused. A reference article with few page views may ground thousands of AI answers, yet a traditional dashboard would recommend retiring it.
- Traffic can signal confusion, not success. An outdated guide may get heavy traffic precisely because customers are stuck.
- Strategic investments stay invisible. Content that powers model grounding gets no credit.
The core idea
QuadCatcher scores every piece of content on two axes, each normalized to a 0–100 scale so different signals can be compared fairly:
- Human Engagement Score (HES): value delivered directly to readers, such as views, return visits, engagement time, and completion.
- Agentic Retrieval Score (ARS): value delivered through AI systems, such as retrieval events, citations, and how often a document actually contributed to an answer.
Plotting the two scores sorts a content portfolio into four quadrants, each with a clear action:
- 👑 Crown Jewels (high human, high agentic): protect and maintain aggressively.
- 💡 Hidden Gems (low human, high agentic): preserve, and optimize metadata and chunking for retrieval.
- ❤️ Human Favorites (high human, low agentic): improve structure and semantic clarity so agents can use them too.
- 🔍 Review Candidates (low human, low agentic): investigate, consolidate, or retire.
A key design decision: freshness isn’t an axis
An article can be in high demand and out of date at the same time. If freshness were an axis, a stale but heavily used article would drop into a different quadrant, hiding the very demand signal that makes it worth fixing. Demand and quality are different dimensions.
So freshness is a quality modifier instead. It colors each point on the chart and feeds an Opportunity Score: content with high retrieval and low freshness rises to the top of the update list. The right action for stale, high-demand content is to update it, not retire it.
Design principles
- Audience agnostic: measure both humans and agents.
- Comparable: normalize different signals to a common scale.
- Explainable: PMs and content teams must be able to understand every score.
- Actionable: every score should point to an action.
- Extensible: new signals can be added later.
What’s new about it
Human engagement scoring is something every analytics platform already does. QuadCatcher’s contribution is pairing it with an agentic retrieval score, so a team can value its content portfolio for both audiences at once. The quadrant chart is the mechanism; the dual-audience model is the idea.
Where it goes next
The roadmap moves from descriptive analytics (what happened?) to prescriptive (what should we do?), then predictive (what will happen?), and eventually to AI-assisted portfolio management that flags update priorities, retirement candidates, and missing content.
