[Paper] Process Analytics -- Data-driven Business Process Management

Source: arXiv - 2512.20703v1

Overview

The paper revisits data‑driven business process analysis—often lumped together under the buzzword process mining—and argues that this narrow view overlooks the social and organisational layers that shape how processes are understood and improved. By introducing the broader concept of process analytics, the authors blend technical analysis techniques with the realities of people, roles, and organisational structures, offering a more holistic roadmap for companies that want to turn event logs into actionable change.

Key Contributions

• Conceptual framework for “process analytics” that extends traditional process mining to include socio‑technical dimensions (stakeholders, governance, culture).
• Inductive‑deductive methodology that derives the framework from literature review, expert interviews, and a real‑world case study.
• Multi‑dimensional taxonomy covering data, techniques, organisational context, and stakeholder involvement.
• Practical illustration through a large‑enterprise case where data‑driven analysis was combined with automation and change‑management practices.
• Critical reflection on the risk of “technical myopia” in current BPM research and practice.

Methodology

1. Literature Mapping (Deductive) – The authors surveyed core BPM and process mining publications to identify existing definitions and gaps.
2. Qualitative Grounding (Inductive) – Semi‑structured interviews with analysts, managers, and IT staff from a multinational firm were conducted to capture lived experiences of process analysis projects.
3. Synthesis – Findings from both strands were merged to build the process analytics model, iteratively refined through workshops with the case‑company’s stakeholders.
4. Validation – The model was applied to a real‑life automation initiative, allowing the authors to observe how the added socio‑technical lenses affected decision‑making and outcomes.

Results & Findings

• Technical focus alone is insufficient: Projects that ignored stakeholder motivations or governance structures suffered from low adoption and missed optimization opportunities.
• Stakeholder‑centric analytics improve ROI: When analysts explicitly mapped who would be affected by a change and involved them early, the subsequent automation yielded a 22 % higher process efficiency gain compared to a “tech‑only” rollout.
• Iterative feedback loops matter: Embedding continuous monitoring and human feedback into the analytics pipeline helped the organization adapt the automated solution as business rules evolved.
• A unified taxonomy clarifies responsibilities: The proposed dimensions (Data, Techniques, Organisation, Stakeholders) served as a checklist that reduced miscommunication between data scientists and business units.

Practical Implications

• For Developers: Build analytics platforms that expose not just raw event logs but also metadata about owners, decision rights, and compliance constraints. APIs should allow easy integration of stakeholder feedback (e.g., via annotation services).
• For BPM Tool Vendors: Offer out‑of‑the‑box modules for stakeholder mapping, governance workflow, and change‑impact simulation alongside traditional process discovery and conformance checking.
• For Enterprises: Adopt a cross‑functional “process analytics office” that includes data engineers, process analysts, and business representatives to co‑design analytics pipelines. This reduces the risk of deploying automation that clashes with existing work practices.
• For Project Managers: Use the taxonomy as a project charter template—explicitly state the data sources, analytical techniques, organisational context, and stakeholder engagement plan before kicking off a mining/automation effort.

Limitations & Future Work

• Single‑case focus: The empirical validation relies on one large organization, which may limit generalisability to SMEs or highly regulated sectors.
• Depth of technical evaluation: While the socio‑technical lens is explored thoroughly, the paper does not benchmark specific mining algorithms against each other.
• Future directions suggested include:
  1. Extending the framework to multi‑organizational ecosystems (e.g., supply chains).
  2. Developing quantitative metrics for the “social” dimensions.
  3. Testing the model across diverse industries to refine its universality.

Authors

• Matthias Stierle
• Karsten Kraume
• Martin Matzner

Paper Information

• arXiv ID: 2512.20703v1
• Categories: cs.SE, cs.ET
• Published: December 23, 2025
• PDF: Download PDF