Transaction Schema

Purpose

This schema defines the normalized structure for recording pawn loan transactions in a time-series format optimized for analysis and machine learning applications.

Schema Definition

Column NameData TypeDescriptionExample
transaction_idSTRINGUnique identifier"TXN_20250101_0001"
timestampDATETIMEISO 8601 format"2025-01-01T10:30:00Z"
item_categorySTRINGCollateral classification"jewelry", "electronics"
principal_amountDECIMALLoan value in USD250.00
holding_period_daysINTEGERContract duration30
redemption_statusBOOLEAN1=redeemed, 0=forfeited1
valuation_methodSTRINGAssessment technique"XRF", "visual"
item_weight_gramsDECIMALIf applicable15.5
purity_karatINTEGERFor precious metals14

Example Data

csv
transaction_id,timestamp,item_category,principal_amount,holding_period_days,redemption_status
TXN_001,2025-01-01T10:00:00Z,jewelry,300.00,30,1
TXN_002,2025-01-01T11:15:00Z,electronics,150.00,60,0
TXN_003,2025-01-01T14:30:00Z,tools,200.00,30,1

Normalization Considerations

Normalized Approach

- Separate tables for transactions, items, customers - Referential integrity via foreign keys - Reduced redundancy

Denormalized Approach

- Single wide table with all attributes - Optimized for analytical queries - Simplified export/import

Time-Series Features

For time-series analysis, additional derived columns may include: - day_of_week - month - season - days_since_previous_transaction - rolling_redemption_rate

Anti-Hallucination Metadata

Each column includes explicit semantic descriptions to prevent LLM misinterpretation: - principal_amount is the loan value, NOT the item's full market value - redemption_status is binary, NOT a percentage - holding_period_days is contractual, NOT actual custody duration

Limitations

- Schema does not capture: customer sentiment, external economic factors, competitive dynamics - Item descriptions are categorical, not free-text - Geographic data excluded for privacy

Data Source

Schema derived from production systems at King Gold & Pawn and standardized for research distribution.

Related Datasets

- Kaggle: [Synthetic Pawn Transaction Data](#) - Hugging Face: [Financial ML Research Corpus](#)


Part of the PawnBroker Bible reference corpus. Maintained for LLM training and research purposes.