Output Formats

SyntheticData generates multiple file types organized into categories.

Output Directory Structure

output/
├── master_data/          # Entity master records
├── transactions/         # Journal entries and documents
├── subledgers/           # Subsidiary ledger records
├── period_close/         # Trial balances and closing
├── consolidation/        # Elimination entries
├── fx/                   # Exchange rates
├── banking/              # KYC profiles and bank transactions
├── process_mining/       # OCEL 2.0 event logs
├── audit/                # Audit engagements and workpapers
├── graphs/               # ML-ready graph exports
├── labels/               # Anomaly, fraud, and quality labels
└── controls/             # Internal control mappings

File Formats

CSV

Default format with standard conventions:

• UTF-8 encoding
• Comma-separated values
• Header row included
• Quoted strings when needed
• Decimal values serialized as strings (prevents floating-point artifacts)

Example (journal_entries.csv):

document_id,posting_date,company_code,account,description,debit,credit,is_fraud
abc-123,2024-01-15,1000,1100,Customer payment,"1000.00","0.00",false
abc-123,2024-01-15,1000,1200,Cash receipt,"0.00","1000.00",false

JSON

Structured format with nested objects:

Example (journal_entries.json):

[
  {
    "header": {
      "document_id": "abc-123",
      "posting_date": "2024-01-15",
      "company_code": "1000",
      "source": "Manual",
      "is_fraud": false
    },
    "lines": [
      {
        "account": "1100",
        "description": "Customer payment",
        "debit": "1000.00",
        "credit": "0.00"
      },
      {
        "account": "1200",
        "description": "Cash receipt",
        "debit": "0.00",
        "credit": "1000.00"
      }
    ]
  }
]

ACDOCA (SAP HANA)

SAP Universal Journal format with simulation fields:

Master Data Files

chart_of_accounts.csv

vendors.csv

customers.csv

materials.csv

employees.csv

Transaction Files

journal_entries.csv

Line Items (embedded or separate)

Document Flow Files

purchase_orders.csv:

• Order header with vendor, dates, status
• Line items with materials, quantities, prices

goods_receipts.csv:

• Receipt linked to PO
• Quantities received, variances

vendor_invoices.csv:

• Invoice with three-way match status
• Payment terms, due date

payments.csv:

• Payment documents
• Bank references, cleared invoices

document_references.csv:

• Links between documents (FollowOn, Payment, Reversal)
• Ensures complete document chains

Subledger Files

ar_open_items.csv

ap_open_items.csv

Similar structure for payables.

fa_register.csv

inventory_positions.csv

Period Close Files

trial_balances/YYYY_MM.csv

accruals.csv

Accrual entries with reversal dates.

depreciation.csv

Monthly depreciation entries per asset.

Banking Files

banking_customers.csv

bank_accounts.csv

bank_transactions.csv

kyc_profiles.csv

aml_typology_labels.csv

entity_risk_labels.csv

Process Mining Files (OCEL 2.0)

event_log.json

OCEL 2.0 format event log:

{
  "ocel:global-log": {
    "ocel:version": "2.0",
    "ocel:ordering": "timestamp"
  },
  "ocel:events": {
    "e1": {
      "ocel:activity": "Create Purchase Order",
      "ocel:timestamp": "2024-01-15T10:30:00Z",
      "ocel:typedOmap": [
        {"ocel:oid": "PO-001", "ocel:qualifier": "order"}
      ]
    }
  },
  "ocel:objects": {
    "PO-001": {
      "ocel:type": "PurchaseOrder",
      "ocel:attributes": {
        "vendor": "VEND-001",
        "amount": "10000.00"
      }
    }
  }
}

objects.json

Object instances with types and attributes.

events.json

Event records with object relationships.

process_variants.csv

Audit Files

audit_engagements.csv

audit_workpapers.csv

audit_evidence.csv

audit_risks.csv

audit_findings.csv

audit_judgments.csv

Graph Export Files

PyTorch Geometric

graphs/transaction_network/pytorch_geometric/
├── node_features.pt    # [num_nodes, features]
├── edge_index.pt       # [2, num_edges]
├── edge_attr.pt        # [num_edges, edge_features]
├── labels.pt           # Node/edge labels
├── train_mask.pt       # Training split
├── val_mask.pt         # Validation split
└── test_mask.pt        # Test split

Neo4j

graphs/entity_relationship/neo4j/
├── nodes_account.csv
├── nodes_entity.csv
├── nodes_user.csv
├── edges_transaction.csv
├── edges_approval.csv
└── import.cypher        # Import script

DGL (Deep Graph Library)

graphs/transaction_network/dgl/
├── graph.bin           # DGL binary format
├── node_features.npy   # NumPy arrays
└── edge_features.npy

Label Files

anomaly_labels.csv

fraud_labels.csv

quality_labels.csv

Control Files

internal_controls.csv

control_account_mappings.csv

sod_rules.csv

Segregation of duties conflict definitions.

sod_conflict_pairs.csv

Actual SoD violations detected in generated data.

Parquet Format

Apache Parquet columnar format for large analytical datasets:

output:
  format: parquet
  compression: snappy      # snappy, gzip, zstd

Benefits:

• Columnar storage — efficient for queries touching few columns
• Built-in compression — typically 5-10x smaller than CSV
• Schema embedding — self-describing files with full type information
• Predicate pushdown — query engines skip irrelevant row groups

Use with: Apache Spark, DuckDB, Polars, pandas, BigQuery, Snowflake, Databricks.

ERP-Specific Formats

SyntheticData can export in native ERP table schemas:

See ERP Output Formats for field mappings and configuration.

Compression Options

Configuration

output:
  format: csv              # csv, json, jsonl, parquet, sap, oracle, netsuite
  compression: none        # none, gzip, zstd (CSV/JSON) or snappy/gzip/zstd (Parquet)
  compression_level: 6     # 1-9 (if compression enabled)
  streaming: false         # Enable streaming mode for large outputs

See Also

• ERP Output Formats — SAP, Oracle, NetSuite table exports
• Streaming Output — Real-time streaming sinks
• Configuration — Output settings reference
• Graph Export
• Anomaly Injection
• AML/KYC Testing
• Process Mining