Auditing Catalog Completeness and Automating Governance for 10,000 SKUs via Explainable AI

Executive P&L Snapshot

Revenue Found

Audit-Proven Search Readiness

Expanded attribute schema depth by 43% across 10,000 pilot SKUs, positioning products for higher organic search discoverability and buyer conversion.

Cost Removed

Automated Merchant Data Cleansing

Replaced manual data remediation queues across multi-language vendor feeds with autonomous AI curation.

Risk Mitigated

Eliminated “Black Box” AI Risk

Deployed Explainable AI (XAI) confidence scores and source citations, giving operators 1-click approval governance over AI decisions.


The Operational Bottleneck: Incomplete Attributes and High Merchant Error Rates

Operating across multiple countries and languages, this $18B GMV marketplace faced significant catalog governance challenges across its merchant network. Merchant data submissions routinely arrived via unstructured spreadsheets, introducing structural quality gaps that harmed search filters and customer experience.

Severe Attribute Incompleteness: A baseline catalog audit revealed a 55% average attribute incompleteness rate, leaving product pages sparse and unoptimized for consumer search filters.

High Data Defect Rate: Merchant inputs suffered from a persistent 5% to 7% data error rate, including misclassified categories, erroneous specifications, and non-compliant main images.

Onboarding Drag: Manual ingestion and translation workflows delayed merchant batch setups, taking up to 14 business days per supplier cohort.


The Fegmo Solution: Image Selection, Attribute Enrichment, and XAI Governance

Fegmo executed a target-paid engagement across 10,000 seller-submitted SKUs, deploying specialized AI agents to analyze media quality, complete attribute schemas, and flag catalog errors.

Unstructured Merchant Feeds → Fegmo Category & Image Agents → XAI Governed Catalog

Main Image Recommendation: Fegmo’s Image Agents analyzed seller-submitted media, automatically scoring and selecting compliant main images to replace low-quality merchant thumbnails.

Schema Expansion and Completion: Autonomous agents extracted missing metadata from raw product files, expanding attribute schema coverage by 43% and raising attribute completion rates across all 10,000 products.

Explainable AI Trust Layer: Every AI-generated attribute recommendation and image tag included explicit confidence scores and source citations. Merchandising operators validated complex listings using simple 1-click approval interfaces without manual data entry.


Quantified Impact

Performance Metric

Baseline (Manual Merchant Ingestion)

Fegmo Agentic OS (10K SKU)

Measured Business Outcome

Catalog Batch Onboarding

14 Business Days

Under 5 Minutes per batch

99%+ Acceleration in Ingestion Speed

Attribute Schema Coverage

55% Average Incompleteness

+43% Schema Expansion

Dramatically Improves Attribute Completion Rates

Ingestion Defect Rate

5% to 7% Data Error Rate

0% Ingestion Defects (Defects Flagged)

Eliminates Incorrect Listings Prior to Syndication


Strategic Takeaway: Proof of Autonomous Decisioning with Human Governance

By executing a focused paid engagement on a 10,000 SKU assortment, Fegmo demonstrated how agentic infrastructure cleanses chaotic supplier feeds while maintaining 100% human-in-the-loop auditability.

The inclusion of Explainable AI confidence scores solved the core enterprise friction point around AI trust, establishing a clear technical foundation for expanding agentic workflows across broader marketplace categories.

Abhishek Ojha