Classification

Classification is the core analytical engine of SpendCraft. It systematically organizes raw transaction data into a standardized, hierarchical structure, converting individual purchase lines into actionable spend categories.

🎯 The Goal of Classification

The primary objective is to assign every line item in a Dataset to a precise, multi-level category path defined by your Taxonomy.

This standardization allows you to compare spend across different business units and geographies using a single, unified lens.

🤖 The Classification Process

Classification is an automated workflow driven by the AI Agent:

1. AI Agent Selection: You select a pre-trained Agent containing rules learned from previously verified data.
2. Field Analysis: The Agent analyzes textual fields like Vendor Name, Transaction Description, and GL Codes.
3. Pattern Mapping: The Agent maps the transaction to the most appropriate taxonomy category based on its training.
4. Confidence Scoring: Each result receives a score. Low-confidence classifications (e.g., below 80%) are flagged for manual audit.

🔄 Continuous Learning Loop

Classification accuracy improves over time through a built-in feedback mechanism:

• Auditing & Correction: Analysts review the results. If a "Laptop" was incorrectly classified as "Office Supplies," they manually correct it to "IT Hardware."
• Model Refinement: SpendCraft captures these corrections and feeds them back into the AI Agent's training data.
• Performance Impact: The Agent updates its internal logic, ensuring it handles similar future transactions correctly without human intervention.

📈 Strategic Importance

• Benchmarking: Compare category spend accurately against industry standards.
• Category Management: Provide managers with a 360-degree view of total addressable spend (e.g., all "Cloud Hosting" spend regardless of the vendor).
• Data Consistency: Ensure every department uses the same definitions and language for expenses.

Related Guides