Column Mapping

Review and approve AI-suggested column mappings before running compliance scans

Yggdrasil requires explicit approval of column mappings before running scans. This ensures transparent, auditable data transformations.

Why Explicit Approval?

This is a core design decision:

Transparent mapping principle: No data transformations happen behind the scenes. You must explicitly confirm which CSV columns map to which compliance schema fields.

Benefits:

• Auditability: Every scan has a documented mapping configuration
• Error prevention: Catch incorrect mappings before generating false violations
• Control: Override AI suggestions when you know your data better

Standard Schema Fields

Yggdrasil maps your CSV columns to these standard fields:

Field	Description	Example CSV Columns
amount	Monetary value	Amount, Transaction_Amount, Payment_Value
step	Time reference	Timestamp, Date, Step, Period
account	Primary entity	Account_ID, Sender, User_ID, Employee_ID
recipient	Secondary entity	Recipient, Destination, To_Account
type	Category/classification	Transaction_Type, Event_Type, Category

Not all fields are required. Only map fields that exist in your dataset and are relevant to your compliance rules.

Mapping Review Process

1. Review AI suggestions

   After upload, Yggdrasil shows suggested mappings with confidence scores:

   {
     "amount": "Transaction_Amount",  // 98% confidence
     "account": "Sender_ID",          // 95% confidence
     "type": "TX_Type"                // 87% confidence
   }
   

   High confidence (>90%) usually means the AI detected a known dataset format.

2. Adjust mappings if needed

   Click any field to change its CSV column mapping:

   • Select a different column from the dropdown
   • Leave unmapped if not applicable
   • The system shows a preview of sample values for verification
3. Set temporal scale

   If your rules use time windows, specify how to interpret your time column:

   • 1.0 = 1 step = 1 day (default for GENERIC datasets)
   • 24.0 = 1 step = 1 hour (default for IBM AML / PaySim)

   Example: If your CSV has step values 1, 2, 3 and scale is 24.0, the engine treats step 1→2 as a 24-hour window.

4. Confirm mapping

   Click “Confirm Mapping” to save your configuration. This generates a mapping_id and marks the audit ready to scan.

Mapping Storage

Confirmed mappings are stored in-memory with the scan configuration:

{
  "mapping_id": "a1b2c3d4-...",
  "upload_id": "e5f6g7h8-...",
  "mapping_config": {
    "amount": "Transaction_Amount",
    "account": "Sender_ID",
    "type": "TX_Type"
  },
  "temporal_scale": 24.0
}

This mapping is applied during the scan and saved with the scan results for auditability.

What Happens During Confirmation

When you click “Confirm Mapping”:

1. Validation: System checks that at least one field is mapped
2. Storage: Mapping config is saved with a unique mapping_id
3. Readiness check: Scan button becomes enabled
4. Response: API returns { mapping_id, ready_to_scan: true }

The mapping is stored in-memory, not persisted to the database. If you refresh the page, you’ll need to re-upload and re-confirm.

Advanced: Cross-Field Comparisons

Some compliance rules compare two fields (e.g., “sender balance != recipient balance”). For these rules, both fields must be mapped:

{
  "conditions": {
    "field": "sender_balance",
    "operator": "not_equals",
    "value": "receiver_balance",
    "value_type": "field"  // Indicates value is another column
  }
}

If your CSV doesn’t have the required columns, you can:

• Disable the rule (toggle it off in the Rules screen)
• Add the missing columns to your CSV and re-upload

Common Mapping Scenarios

Scenario 1: GDPR Consent Audit

You’re auditing consent records. Your CSV has:

• data_subject_id → map to account
• consent_date → map to step
• No monetary values → leave amount unmapped

Scenario 2: AML Transaction Monitoring

You’re scanning financial transactions. Your CSV has:

• transaction_amount → map to amount
• timestamp → map to step
• sender_account → map to account
• receiver_account → map to recipient
• transaction_type → map to type

Scenario 3: SOC2 Access Logs

You’re auditing system access. Your CSV has:

• user_id → map to account
• login_time → map to step
• action → map to type
• No amounts or recipients → leave unmapped

Type Coercion

CSV files produce string values. The engine automatically coerces types:

• "true" / "false" → boolean
• "16" / "99.5" → numeric
• Date strings → parsed as timestamps (if used in temporal rules)

You don’t need to pre-format your CSV — Yggdrasil handles type conversion during scan execution.

Next Steps

After confirming your mapping:

1. Optionally run PII detection → PII Detection
2. Review rules and toggle any you want to disable
3. Run the compliance scan → Compliance Scanning