At the core, this wasn’t a UI problem. This was a trust collapse.
Enterprise clients were managing 100,000+ SKUs across global retailers.
But the platform meant to streamline that data was riddled with inconsistencies, duplications, and outdated fields — pulled from a web of disconnected sources (flat files, PIMs, internal systems, retailer APIs).
Users didn’t know what content was correct, what was live, or what would overwrite what.
So they stopped using the platform entirely — and started doing everything manually:
Spreadsheets
Screenshots
Email threads
Risky custom scripts
Every edit was a gamble, every sync a potential disaster.
The cost? Hours of manual effort, millions in lost revenue, and a system no one wanted to touch.
The business saw the symptoms — lack of adoption, support tickets piling up, missed SLAs — but hadn’t yet connected the root problem: Users didn’t trust the platform enough to use it.
✅ THE SOLUTION
What we built — and how it fixed the trust gap.
I designed a unified content operations system with two integrated modules:
1. Content Intelligence System
To help users validate, clean, and publish accurate content at scale:
AI-powered discrepancy detection flagged vague, conflicting, or incomplete data
Confidence scores guided where to act first
Inline editing with real-time validation made fixes fast and visible
Versioning and history restored accountability and reduced fear of breaking things
2. Variant Structuring System
To fix broken product relationships and restore logical product grouping:
AI-suggested groupings based on attribute logic (color, size, etc.)
Visual dashboards showed bloat, errors, and sales impact
Side-by-side comparison views enabled fast manual restructuring
Validated groupings were synced into downstream systems reliably
Key Design Strategies:
Dashboard-first model: Prioritized action, not noise
Explainable AI: Users could see why suggestions were made — and override with confidence
Cross-functional alignment: Built ingestion logic with engineering, training data with AI/ML, and validation flows with customer success
Designed for imperfection: The system handled bad data, not just ideal cases
🎯 THE OUTCOME
What changed — and why it mattered.
Time to identify content issues dropped from ~1 hour to <10 minutes
Resolution time in pilot clients improved by 50%
Support tickets forecasted to decrease by 30–40%
Platform adoption increased significantly, as users began to trust and rely on it again
System now serves as the foundation for future automation and compliance tools
.png)
.004.jpeg)
.007.jpeg)
.010.jpeg)
.017.jpeg)
