Industry-Specific Features

SyntheticData includes industry-specific transaction modeling with authentic terminology, master data structures, and anomaly patterns. Three industries have full generator implementations (Manufacturing, Retail, Healthcare), while three additional industries (Technology, Financial Services, Professional Services) are available as configuration presets with industry-appropriate GL structures and anomaly rates.

Overview

Each industry module provides:

Industry-specific transactions: Authentic transaction types using correct terminology
Master data structures: Industry-specific entities (BOM, routings, clinical codes, etc.)
Anomaly patterns: Industry-authentic fraud and error patterns
GL account structures: Industry-appropriate chart of accounts
Configuration options: Fine-grained control over industry characteristics

Implementation Status

Industry	Status	Transaction Types	Master Data	Anomaly Patterns	Benchmarks
Manufacturing	Full generator	13 types	BOM, routings, work centers	5 patterns	Yes
Retail	Full generator	11 types	Stores, POS, loyalty	6 patterns	Yes
Healthcare	Full generator	9 types	ICD-10, CPT, DRG, payers	6 patterns	Yes
Technology	Config preset	Config-only	—	3 anomaly rates	Yes
Financial Services	Config preset	Config-only	—	3 anomaly rates	Yes
Professional Services	Config preset	Config-only	—	3 anomaly rates	No

Full generator industries have dedicated Rust enum types with per-transaction generation logic, dedicated master data structures, and industry-specific anomaly injection. Config preset industries use the standard generator pipeline but apply industry-appropriate GL account structures, transaction distributions, and anomaly rates through configuration.

Manufacturing

Transaction Types

#![allow(unused)]
fn main() {
pub enum ManufacturingTransaction {
    // Production
    WorkOrderIssuance,      // Create production order
    MaterialRequisition,    // Issue materials to production
    LaborBooking,           // Record labor hours
    OverheadAbsorption,     // Apply manufacturing overhead
    ScrapReporting,         // Report production scrap
    ReworkOrder,            // Create rework order
    ProductionVariance,     // Record variances

    // Inventory
    RawMaterialReceipt,     // Receive raw materials
    WipTransfer,            // Transfer between work centers
    FinishedGoodsTransfer,  // Move to finished goods
    CycleCountAdjustment,   // Inventory adjustments

    // Costing
    StandardCostRevaluation,  // Update standard costs
    PurchasePriceVariance,    // Record PPV
}
}

Master Data

manufacturing:
  bom:
    depth: 4                    # BOM levels (3-7 typical)
    yield_rate: 0.97            # Expected yield
    scrap_factor: 0.02          # Scrap percentage
  routings:
    operations_per_product: 5   # Average operations
    setup_time_minutes: 30      # Default setup time
  work_centers:
    count: 20
    capacity_hours: 8
    efficiency: 0.85

Anomaly Patterns

Anomaly	Description	Detection Method
Yield Manipulation	Reported yield differs from actual	Variance analysis
Labor Misallocation	Labor charged to wrong order	Cross-reference
Phantom Production	Production orders with no output	Data analytics
Obsolete Inventory	Aging inventory not written down	Aging analysis
Standard Cost Manipulation	Inflated standard costs	Trend analysis

Configuration

industry_specific:
  enabled: true
  manufacturing:
    enabled: true
    bom_depth: 4
    just_in_time: false
    production_order_types:
      - standard
      - rework
      - prototype
    quality_framework: iso_9001
    supplier_tiers: 2
    standard_cost_frequency: quarterly
    target_yield_rate: 0.97
    scrap_alert_threshold: 0.03
    anomaly_rates:
      yield_manipulation: 0.005
      labor_misallocation: 0.008
      phantom_production: 0.002
      obsolete_inventory: 0.01

Retail

Transaction Types

#![allow(unused)]
fn main() {
pub enum RetailTransaction {
    // Point of Sale
    PosSale,                // Register sale
    ReturnRefund,           // Customer return
    VoidTransaction,        // Voided sale
    EmployeeDiscount,       // Staff discount
    LoyaltyRedemption,      // Points redemption

    // Inventory
    InventoryReceipt,       // Receive from DC
    StoreTransfer,          // Store-to-store
    MarkdownRecording,      // Price reductions
    ShrinkageAdjustment,    // Inventory loss

    // Cash Management
    CashDrop,               // Safe deposit
    RegisterReconciliation, // Drawer count
}
}

Store Types

retail:
  stores:
    types:
      - flagship      # High-volume, full assortment
      - standard      # Normal retail store
      - express       # Small format, convenience
      - outlet        # Discount/clearance
      - warehouse     # Bulk/club format
      - pop_up        # Temporary locations
      - digital       # E-commerce only

Anomaly Patterns

Anomaly	Description	Detection Method
Sweethearting	Not scanning items	Video analytics
Skimming	Cash theft from register	Cash variance
Refund Fraud	Fake returns	Return pattern
Receiving Fraud	Short shipment theft	3-way match
Coupon Fraud	Invalid coupon use	Coupon validation
Employee Discount Abuse	Unauthorized discounts	Policy review

Configuration

industry_specific:
  enabled: true
  retail:
    enabled: true
    store_types:
      - standard
      - express
      - outlet
    shrinkage_rate: 0.015
    return_rate: 0.08
    markdown_frequency: weekly
    loss_prevention:
      camera_coverage: 0.85
      eas_enabled: true
    pos_anomaly_rates:
      sweethearting: 0.002
      skimming: 0.001
      refund_fraud: 0.003

Healthcare

Transaction Types

#![allow(unused)]
fn main() {
pub enum HealthcareTransaction {
    // Revenue Cycle
    PatientRegistration,    // Register patient
    ChargeCapture,          // Record charges
    ClaimSubmission,        // Submit to payer
    PaymentPosting,         // Record payment
    DenialManagement,       // Handle denials

    // Clinical
    ProcedureCoding,        // CPT codes
    DiagnosisCoding,        // ICD-10 codes
    SupplyConsumption,      // Medical supplies
    PharmacyDispensing,     // Medications
}
}

Coding Systems

healthcare:
  coding:
    icd10: true         # Diagnosis codes
    cpt: true           # Procedure codes
    drg: true           # Diagnosis Related Groups
    hcpcs: true         # Supplies/equipment

Payer Mix

healthcare:
  payer_mix:
    medicare: 0.40
    medicaid: 0.20
    commercial: 0.30
    self_pay: 0.10

Compliance Frameworks

healthcare:
  compliance:
    hipaa: true           # Privacy rules
    stark_law: true       # Physician referrals
    anti_kickback: true   # AKS compliance
    false_claims_act: true

Anomaly Patterns

Anomaly	Description	Detection Method
Upcoding	Higher-level code than justified	Code validation
Unbundling	Splitting bundled services	Bundle analysis
Phantom Billing	Billing for unrendered services	Audit
Duplicate Billing	Same service billed twice	Duplicate check
Kickbacks	Physician referral payments	Relationship analysis
HIPAA Violations	Unauthorized data access	Access logs

Configuration

industry_specific:
  enabled: true
  healthcare:
    enabled: true
    facility_type: hospital  # hospital, physician_practice, etc.
    payer_mix:
      medicare: 0.40
      medicaid: 0.20
      commercial: 0.30
      self_pay: 0.10
    coding_system:
      icd10: true
      cpt: true
      drg: true
    compliance:
      hipaa: true
      stark_law: true
      anti_kickback: true
    avg_daily_encounters: 200
    avg_charges_per_encounter: 8
    anomaly_rates:
      upcoding: 0.02
      unbundling: 0.015
      phantom_billing: 0.005
      duplicate_billing: 0.008

Technology

Transaction Types

License revenue recognition
Subscription billing
Professional services
R&D capitalization
Deferred revenue

Configuration

industry_specific:
  enabled: true
  technology:
    enabled: true
    revenue_model: subscription  # license, subscription, usage
    subscription_revenue_percent: 0.70
    professional_services_percent: 0.20
    license_revenue_percent: 0.10
    r_and_d_capitalization_rate: 0.15
    deferred_revenue_months: 12
    anomaly_rates:
      premature_revenue: 0.008
      channel_stuffing: 0.003
      improper_capitalization: 0.005

Financial Services

Transaction Types

Loan origination
Interest accrual
Fee income
Trading transactions
Customer deposits
Wire transfers

Configuration

industry_specific:
  enabled: true
  financial_services:
    enabled: true
    institution_type: commercial_bank
    regulatory_framework: us  # us, eu, uk
    loan_portfolio_size: 1000
    avg_loan_amount: 250000
    loan_loss_provision_rate: 0.02
    fee_income_percent: 0.15
    trading_volume_daily: 50000000
    anomaly_rates:
      loan_fraud: 0.003
      trading_fraud: 0.001
      account_takeover: 0.002

Professional Services

Transaction Types

Time and billing
Engagement management
Trust account transactions
Expense reimbursement
Partner distributions

Configuration

industry_specific:
  enabled: true
  professional_services:
    enabled: true
    billing_model: hourly  # hourly, fixed_fee, contingency
    avg_hourly_rate: 350
    utilization_target: 0.75
    realization_rate: 0.92
    trust_accounting: true
    engagement_types:
      - audit
      - tax
      - advisory
      - litigation
    anomaly_rates:
      billing_fraud: 0.004
      trust_misappropriation: 0.001
      expense_fraud: 0.008

Industry Benchmarks

SyntheticData provides pre-configured ML benchmarks for each industry:

#![allow(unused)]
fn main() {
// Get industry-specific benchmark
let bench = get_industry_benchmark(IndustrySector::Healthcare);

// Available benchmarks
let manufacturing = manufacturing_fraud_5k();
let retail = retail_fraud_10k();
let healthcare = healthcare_fraud_5k();
let technology = technology_fraud_3k();
let financial = financial_services_fraud_5k();
}

Benchmark Features

Each industry benchmark includes:

Industry-specific transaction features
Relevant anomaly types
Appropriate cost matrices
Industry-specific evaluation metrics

Best Practices

Match industry to use case: Select the industry that matches your target domain
Use industry presets first: Start with default settings before customizing
Enable industry-specific anomalies: These provide realistic fraud patterns
Consider regulatory context: Enable compliance frameworks relevant to your industry
Use industry benchmarks: Evaluate models against industry-specific baselines

Output Files

File	Description
`industry_transactions.csv`	Industry-specific transaction log
`industry_master_data.json`	Industry-specific entities
`industry_anomalies.csv`	Industry-specific anomaly labels
`industry_gl_accounts.csv`	Industry-specific chart of accounts

Keyboard shortcuts

SyntheticData Documentation