Catalog quality — findings not fill rates
Catalog Quality in the Agent Control Center measures catalog health from open agent insights and pending approvals — not from PIM attribute fill rate or legacy completeness scores alone.
This aligns with how modern Agentic PIM teams operate: quality is what specialist agents find and propose to fix, with human gates before publish.
Two different signals
| Signal | What it measures | Where to find it |
|---|---|---|
| Catalog Quality (ACC) | Count/severity of open insights from specialist agents; pending approvals | Agent Control Center → Catalog Quality |
| Completeness score | % of required attributes filled per family/channel | Product views · Completeness guide |
Both matter — they answer different questions:
- Completeness — “Can we export this SKU to Amazon?” (publish gate)
- Catalog Quality — “What problems do our agents see across the catalog?” (operational intelligence)
A product can show 100% completeness and still have critical SEO or compliance insights from agents. Conversely, high catalog quality (no open insights) does not mean every attribute is filled — it means agents have not flagged known problems.
What Catalog Quality shows
The Catalog Quality page rolls up:
- Overall health score — derived from open insights, pending approvals, and recent failed runs
- Quality dimensions — e.g. content, SEO, taxonomy, compliance, syndication readiness
- Deep links — dimension cards link to Insights filtered by insight type
Data comes from agent run output — not a separate “quality AI” scoring products in isolation.
Scoring caveats
| What it measures | What it does not measure |
|---|---|
| Open insights from specialist agents | PIM attribute fill rate |
| Pending approvals (headline factor) | Whether agents scanned the entire catalog |
| Recent failed runs (headline factor) | Channel export success rate |
No open insights → dimensions may show 100% — meaning no known open problems, not catalog is complete.
The UI banner states: Scores roll up open insights and pending approvals from your specialist agents — not a separate AI agent.
How teams use it
- Daily standup — check headline score and critical insight count
- Before syndication — resolve syndication and compliance insights; use completeness for export gates
- Agent tuning — if one dimension stays red, adjust agent schedules or policies
Resolve insights
Open Insights to triage findings: filter by severity, bulk-resolve when fixed, or convert to approval proposals when agents recommend writes.
→ Agent Control Center · Specialist agents