Back to Documentation
Core Concepts

Domain Intelligence

Domain Intelligence: How AUM Understands Your Data

AUM's revolutionary Domain Intelligence engine is what sets it apart from generic analytics tools.

How It Works

When you upload a file, AUM analyzes:

  1. Column Names: Identifies industry-specific terminology
  2. Data Patterns: Recognizes common data structures (e.g., SKU formats, order IDs)
  3. Value Types: Understands numeric vs. categorical data
  4. Relationships: Detects foreign keys and hierarchies

Multi-Layer Detection

AUM uses 6 sophisticated layers:

  1. File Name Analysis (Weight: 3.0)

    • Scans filename for industry keywords
    • Example: "sales_orders_2024.csv" → E-commerce likelihood increases

  2. Deep Column Analysis (Weight: 2.5)

    • Matches column names against 1000+ industry terms
    • Example: Columns like "sku", "cart_value" → E-commerce

  3. Pattern Recognition (Weight: 4.0)

    • Validates data formats (e.g., ORDER-123456, VIN numbers)
    • High accuracy indicator

  4. Statistical Analysis (Weight: 2.0)

    • Analyzes numeric vs. categorical ratio
    • Manufacturing: >60% numeric columns

  5. Semantic Column Groups (Weight: 3.0)

    • Looks for metric + dimension pairs
    • Example: "revenue" + "customer_id" → Business data

  6. Required Column Check (Weight: 5.0 - CRITICAL)

    • Verifies presence of domain-essential columns
    • Highest confidence boost

Supported Domains

E-commerce 🛒

Detection Keywords: order, sku, cart, product, checkout, conversion, gmv, aov

Key Metrics Generated:

  • Gross Merchandise Value (GMV)
  • Average Order Value (AOV)
  • Conversion Rate
  • Cart Abandonment Rate
  • Return Rate

Sample Insights:

  • "Top 10 SKUs driving 45% of revenue"
  • "Cart abandonment rate increased 15% this week"
  • "Product category 'Electronics' showing 23% growth"

Manufacturing 🏭

Detection Keywords: plant, line, oee, throughput, defect, batch, downtime

Key Metrics Generated:

  • Overall Equipment Effectiveness (OEE)
  • Production Throughput
  • Defect Rate
  • Downtime Hours
  • First Pass Yield

Sample Insights:

  • "Line 3 operating at 78% OEE (below 85% target)"
  • "Defect rate spiked to 3.2% on Night Shift"
  • "Machine M-007 has 12 hours unplanned downtime"

Healthcare 🏥

Detection Keywords: patient, diagnosis, admission, bed, appointment, icd

Key Metrics Generated:

  • Bed Occupancy Rate
  • Average Length of Stay (LOS)
  • Readmission Rate
  • Patient Wait Time
  • Appointment No-Show Rate

Sample Insights:

  • "Cardiology department at 92% bed capacity"
  • "Average LOS increased from 4.2 to 5.1 days"
  • "15% no-show rate for Monday morning slots"

Retail 🏪

Detection Keywords: store, pos, footfall, inventory, transaction

Key Metrics:

  • Footfall vs. Sales Conversion
  • Average Basket Size
  • Inventory Turnover
  • Store-Wise Performance
  • Peak Hours Analysis

Finance 💰

Detection Keywords: loan, npa, account, interest, disbursement

Key Metrics:

  • Non-Performing Asset (NPA) Ratio
  • Loan-to-Deposit Ratio
  • Interest Income
  • Portfolio Quality
  • Default Rate Trends

Automotive 🚗

Detection Keywords: dealer, vin, model, booking, vehicle

Key Metrics:

  • Sales by Dealer
  • Test Drive to Booking Ratio
  • Model-Wise Performance
  • Regional Market Share
  • Inventory Days

Confidence Scoring

AUM assigns a confidence score (0-100%) to domain detection:

  • High Confidence (>70%): Strong match, domain-specific KPIs activated
  • Medium Confidence (40-70%): Likely match, suggests verification
  • Low Confidence (<40%): Falls back to generic analysis

Domain-Specific Advantages

Why Domain Matters

Generic Tool Output:

Column: amount
Average: 1,234.56
Max: 5,000.00

AUM E-commerce Output:

Average Order Value: ₹1,234.56 (+12% vs. last month)
⚠️ ALERT: Top 10% of orders contribute 48% of GMV
💡 INSIGHT: Basket size peaks on weekends (₹1,890 avg)
✅ RECOMMENDATION: Increase mid-week promotions

Switching Domains

On Professional/Enterprise plans, you can manually override domain detection if needed:

  1. Go to Dataset Settings
  2. Click "Change Domain"
  3. Select from available domains
  4. Confirm - metrics will regenerate

API Integration

Access domain intelligence programmatically:

const response = await fetch('/api/data/upload', {
  method: 'POST',
  body: formData
});

const result = await response.json();
console.log(result.domain); // "ecommerce"
console.log(result.confidence); // 0.87
console.log(result.domain_scores); // { ecommerce: 0.87, retail: 0.13 }

Best Practices

  1. Use Descriptive Column Names: "order_date" > "date1"
  2. Include Context in Filenames: "q4_sales_report.csv" vs. "data.csv"
  3. Maintain Consistent Formats: Standard date formats, currency symbols
  4. Remove Test Data: Clean production data yields better detection

Troubleshooting

Q: Why did AUM detect wrong domain? A: Check if your columns use industry-standard naming. You can manually switch domains on Pro/Enterprise plans.

Q: Confidence is only 45%, should I worry? A: Medium confidence still provides useful insights. Consider renaming columns to match industry standards.

Q: Can I add custom domains? A: Enterprise customers can request custom domain training. Contact sales@aumdatalabs.com

Next: Learn about Natural Language Queries

← Previous

Getting Started

Next →

Natural Language Queries