Mining patterns for proactive management of procure-to-pay exceptions

This paper explores a novel approach to addressing exceptions in the procure-to-pay (P2P) process by analyzing recurring patterns that cause procurement inefficiencies. The P2P process, which involves issuing purchase orders, receiving goods, and validating invoices through three-way matching, often encounters exceptions such as pricing discrepancies, quantity mismatches, and missing documentation. These issues can lead to delays, financial errors, and strained supplier relationships. Existing black-box models such as neural networks, while effective at detection of exceptions, lack transparency. On the other hand, traditional white-box methods such as classification trees fall short in identifying complex patterns that drive recurring exceptions. This paper presents an enhanced white-box solution, which improves the generation and prioritization of classification rules by optimizing the depth, coverage, and explainability of patterns. The proposed method allows for more accurate and comprehensive exception management, empowering procurement teams to make informed decisions with greater transparency.