IBM Multilevel Process Mining vs de facto Object-Centric Process Mining approaches

TL;DR

The paper analyzes IBM's Multilevel Process Mining and Object-Centric Process Mining (OCPM), highlighting their strengths and trade-offs. It introduces Organization Mining as a synthesis that combines cross-entity, end-to-end views with object-centric data structures to simplify data preparation and enable enterprise-wide analysis. The work discusses data models (OCEL/OCED), conformance approaches, throughput-time calculations across objects, and cross-process statistics, arguing for a unified approach that improves portability, scalability, and actionable insights. The practical impact is a more integrated, organization-level process intelligence capability that supports faster, more informed decision-making across multi-process environments.

Abstract

The academic evolution of process mining is moving toward object centric process mining, marking a significant shift in how processes are modeled and analyzed. IBM has developed its own distinctive approach called Multilevel Process Mining. This paper provides a description of the two approaches and presents a comparative analysis of their respective advantages and limitations. IBM leveraged this comparison to drive the evolution of IBM Process Mining product, creating the new Organizational Mining feature, an innovation that combines the best of the two approaches. Demonstrate the potential of this novel, innovative and distinct methodology with an example.

Figures (18)

• Figure 1: Example of an event log used in process mining, highlighting mandatory fields and additional attributes.
• Figure 2: Tabular extraction from a Multilevel event log exemplifying typical relationships between objects (Order, Receipt and Invoice).
• Figure 3: Model of a P2P Multilevel process composed by Order, Receipt and Invoice entities in IBM Process Mining.
• Figure 4: Model and statistics of a P2P Multilevel process composed by Order, Receipt and Invoice entities in IBM Process Mining.
• Figure 5: Results of the comparison between data derived and reference model in a P2P Multilevel process composed by Order, Receipt and Invoice entities in IBM Process Mining.