A-Eye QC Enhancement

Hierarchical Rules &
Dimensional Trend Integrity

Core Capabilities

Four pillars of the enhanced QC system

Fixed Schema + Override Rules

4-Level Cascading Override Architecture

Rule IDs are fixed at Global level. Lower levels can only OVERRIDE existing rules - no conflicts possible.

6 Categories: VIS, DAT, FMT, BRD, TXT, LAY
24 Standard Rules with unique IDs
Source tracking shows where each rule comes from
Override chain visible for debugging
Explore Feature

Trend Analysis

Statistical Anomaly Detection

Dimensional trend integrity with seasonality-aware baseline calculations.

Z-Score based anomaly detection
Historical baseline calculation
Retail calendar seasonality factors
Week-over-week variance tracking
Explore Feature

Anomaly Attribution

3-Layer Attribution Model

Explain trend anomalies through statistical analysis, retail calendar, and user-defined events.

Layer 1: Statistical anomaly detection (Z-Score)
Layer 2: Retail calendar + seasonality rules
Layer 3: User-defined contextual events
Attribution: Explained / Partial / Unexplained
Explore Feature

AI Integration

Multi-Provider LLM Support

Flexible AI backend supporting Claude, OpenAI, and Gemini for analysis.

Claude (Primary - Best Reasoning)
OpenAI GPT-4o (Fallback)
Gemini 1.5 Pro (Alternative)
Automatic provider switching
Explore Feature

Fixed Schema + Override Architecture

Rule IDs defined at Global level - lower levels can only override

Level 1
Global
Schema + Defaults
Level 2
Client
Overrides Only
Level 3
Document
Overrides Only
Level 4
Slide
Overrides Only

No Rule Conflicts

Since lower levels can only override existing rule IDs (not create new ones), there are never conflicting rule definitions.

Source Tracking

Every rule shows where its effective value comes from (Global, Client, Document, or Slide) with full override chain history.

Example: DAT_001 (Variance Threshold)

Global: 10%
Walgreens: 5%
Effective: 5% (source: client)

Implementation Journey

From idea to production-ready solution

Phase I

Discovery & Design

  • Identify QC pain points
  • Define rule hierarchy concept
  • Design data architecture
Phase II

Core Engine

  • Build rule resolution engine
  • Implement override logic
  • Create validation APIs
Phase III

Intelligence Layer

  • Integrate AI providers
  • Build anomaly detection
  • Add trend analysis
Phase IV

User Experience

  • Design intuitive interfaces
  • Add visual feedback
  • Enable human-in-the-loop
Phase V

Production Ready

  • Optimize performance
  • Add monitoring
  • Deploy & iterate

AI-Assisted Workflow

Human-in-the-loop event confirmation

Anomaly Detected
LLM Analysis
Human Review
Confirmed Event

Ready to Explore?

Navigate through the enhanced Visual QC system and experience the new capabilities firsthand.

A-Eye QC Enhancement | 5 Sprints | 26 Files | ~5,500 Lines of Code

Based on requirements from Circana