Skip to main content

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

SignalWhat it measuresWhere to find it
Catalog Quality (ACC)Count/severity of open insights from specialist agents; pending approvalsAgent Control Center → Catalog Quality
Completeness score% of required attributes filled per family/channelProduct 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)
Findings, not fill rates

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 measuresWhat it does not measure
Open insights from specialist agentsPIM 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

  1. Daily standup — check headline score and critical insight count
  2. Before syndication — resolve syndication and compliance insights; use completeness for export gates
  3. 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