dr-anomalies

Datarails Anomaly Detection

Automated anomaly detection for Finance OS tables - find data quality issues, outliers, and suspicious patterns.

Workflow

Step 1: Verify Authentication

If any tool call fails with an authentication or connection error, guide the user to connect via the Connectors UI ("+" > Connectors > Datarails > Connect).

Step 2: Run Detection

1. Get table schema for context
2. Run detect_anomalies for comprehensive analysis
3. Optionally run profile_numeric_fields and profile_categorical_fields for deeper stats
4. If specific anomalies need investigation, use get_records_by_filter to fetch examples

Step 3: Present Findings

Organize findings by severity:

• 🔴 Critical - Requires immediate attention
• 🟠 High - Should be addressed soon
• 🟡 Medium - Worth investigating
• 🟢 Low - Minor issues or informational

Arguments

Anomaly Types Detected

Example Interactions

User: "/dr-anomalies 11442"

🔍 Anomaly Detection: GL Transactions (ID: 11442)
═══════════════════════════════════════════════════════════

Scanned 125,432 records | Found 47 anomalies

🔴 CRITICAL (3 findings)
───────────────────────────────────────────────────────────

1. DUPLICATE TRANSACTIONS
   • 23 potential duplicate records detected
   • Same amount, date, and vendor within 1 minute
   • Records: [45231, 45232], [67892, 67893], ...

   💡 Recommendation: Review for accidental double-entry
   📋 Query: /dr-query 11442 "transaction_id IN (45231, 45232)"

2. FUTURE-DATED TRANSACTIONS
   • 5 transactions with posting_date > today
   • Dates range from 2024-02-15 to 2024-12-31

   💡 Recommendation: Verify if these are planned entries
   📋 Query: /dr-query 11442 "posting_date > '2024-01-20'"

3. NEGATIVE INVENTORY QUANTITIES
   • 8 records with quantity < 0
   • Min value: -500 (record 89234)

   💡 Recommendation: Check if returns are properly coded

🟠 HIGH (12 findings)
───────────────────────────────────────────────────────────

4. AMOUNT OUTLIERS
   • 127 transactions beyond normal range
   • 115 above $500,000 (expected max ~$250,000)
   • 12 below -$100,000 (expected min ~-$50,000)

   💡 Recommendation: Verify large transactions are approved

5. HIGH NULL RATE: vendor_name
   • 2,341 records (1.87%) missing vendor_name
   • But vendor_id is present

   💡 Recommendation: Join with vendor master to populate

...

🟡 MEDIUM (18 findings)
───────────────────────────────────────────────────────────

12. RARE CATEGORY VALUES
    • department contains 3 values appearing < 10 times
    • Values: "TEST", "MIGRATION", "UNKNOWN"

    💡 Recommendation: Standardize or reclassify

...

🟢 LOW (14 findings)
───────────────────────────────────────────────────────────

35. TRAILING WHITESPACE
    • account_code has 45 values with trailing spaces

    💡 Recommendation: Trim during ETL

═══════════════════════════════════════════════════════════
📊 SUMMARY
═══════════════════════════════════════════════════════════

| Severity | Count | Action                    |
|----------|-------|---------------------------|
| Critical | 3     | Investigate immediately   |
| High     | 12    | Address this week         |
| Medium   | 18    | Plan for remediation      |
| Low      | 14    | Fix during maintenance    |

Data Quality Score: 87/100 ⚠️
Primary concerns: Duplicates, Future dates, Amount outliers

User: "/dr-anomalies 11442 --severity critical"

🔴 Critical Anomalies: GL Transactions

Found 3 critical issues requiring immediate attention:

1. DUPLICATE TRANSACTIONS (23 records)
   ...

2. FUTURE-DATED TRANSACTIONS (5 records)
   ...

3. NEGATIVE INVENTORY QUANTITIES (8 records)
   ...

User: "/dr-anomalies 11442 --type outliers"

📊 Outlier Analysis: GL Transactions

Analyzed 8 numeric fields for statistical outliers (|z| > 3)

amount: 127 outliers
├── Above 3σ: 115 records
│   ├── Max: $8,750,000 (z=12.4)
│   ├── Sample: [45231: $2.1M], [67892: $1.8M], [89234: $1.5M]
│   └── Pattern: Mostly Q4 entries (82%)
└── Below -3σ: 12 records
    ├── Min: -$1,250,000 (z=-8.2)
    └── Sample: [12345: -$800K], [23456: -$650K]

quantity: 23 outliers
├── All above 3σ (high quantities)
├── Max: 10,000 (z=5.1)
└── 18 of 23 are from department="Warehouse"

unit_cost: 45 outliers
...

Investigation Workflow

When anomalies are detected:

1. Review findings - Understand the scope and patterns
2. Fetch examples - Use /dr-query to see actual records
3. Verify business rules - Some "anomalies" may be valid
4. Document decisions - Note which are false positives
5. Create remediation plan - Prioritize by severity

Related Skills

• /dr-profile - Detailed field statistics
• /dr-query - Fetch specific records
• /dr-tables - Understand table structure