Process Mining Manifesto — IEEE Task Force on Process Mining 2012

The foundational manifesto of the process mining field, authored by the IEEE Task Force on Process Mining (Computational Intelligence Society / IEEE, established 2009) and led by entities/wil-van-der-aalst. Published in BPM Workshops 2011 (LNBIP 99, Springer). Roughly 70 signatories across software vendors (Software AG, Futura Process Intelligence, HP, IBM, Infosys, Fluxicon, Businesscape, Fujitsu, Stereologic, Pallas Athena…), consultancy/end-users (Gartner, Deloitte, Rabobank, BWI Systeme GmbH, Excellentia BPM…), and research institutes (TU/e, TU Vienna, QUT, University of Padova, Politècnica de Catalunya, Bari, Haifa, Bologna, Tartu, Innsbruck, Stevens, Cranfield, K.U. Leuven…). The short Expert Voice companion to Van der Aalst's textbook of the following year.

Summary

Process mining is defined as a young research discipline sitting between computational intelligence / data mining and process modelling / analysis. It aims to discover, monitor, and improve real (not assumed) processes by extracting knowledge from event logs readily available in today's information systems. The manifesto is positioned explicitly as a public declaration of principles and intentions by the IEEE Task Force, targeting software developers, scientists, consultants, business managers, and end-users.

The document is organised in four parts:

State of the Art (§2). Defines event logs (activity + case + timestamp + optional resource / data), positions process mining as the "missing link" between data mining and BPM, and presents the three basic types: discovery (log → model), conformance (log + model → diagnostics), enhancement (log + model → improved/extended model). Four perspectives — control-flow, organisational, case, time — are declared orthogonal to the three types. The text flags three common misconceptions: (1) process mining is not limited to control-flow; (2) it is not just data mining; (3) it is not limited to offline analysis — it supports runtime operational support via detect / predict / recommend. The L* life-cycle of a mining project is sketched in five stages (plan and justify → extract → control-flow model + connect log → integrated multi-perspective model → operational support).

Guiding Principles (§3). Six prescriptions for good process mining practice:

GP1: Event data as first-class citizens. Logs should be trustworthy, complete, have well-defined semantics, be safe (privacy), and referenceable. A five-star maturity scale is tabulated: ★ (paper trails) up to ★★★★★ (semantically annotated logs with ontology alignment).
GP2: Log extraction should be driven by questions. Without questions, extraction from a large ERP database is intractable.
GP3: Concurrency, choice and basic control-flow constructs must be supported. Otherwise discovered models degenerate (compact but over-general, or explicit but exploding in size).
GP4: Events should be related to model elements. The log-model linkage enables replay, conformance, and time annotation.
GP5: Models should be treated as purposeful abstractions. Different stakeholders need different views (strategic / tactical / operational); no "perfect map" exists.
GP6: Process mining should be a continuous process. Models are not one-off artefacts; they should be re-mined as the underlying process drifts.

Challenges (§4). Eleven named challenges frame the research agenda:

C1 Finding, merging, cleaning event data (distribution, object- vs process-centricity, incompleteness, outliers/noise, granularity, context).
C2 Dealing with complex event logs of diverse characteristics (size, variability, imbalance).
C3 Creating representative benchmarks (real + synthetic; evaluation metrics; cross-validation adapted to control-flow).
C4 Dealing with concept drift (splitting logs, footprint analysis; periodicity, seasonality, changing conditions).
C5 Improving representational bias used for discovery (target language choice limits search space).
C6 Balancing quality criteria — fitness, simplicity, precision, generalisation (Occam; avoid flower-models and over-fits; open-world assumption).
C7 Cross-organisational mining (collaborative jigsaw settings and shared-process settings; privacy preservation).
C8 Providing operational support — detect, predict, recommend — at runtime.
C9 Combining process mining with other types of analysis (operations research, simulation, visual analytics).
C10 Improving usability for non-experts (self-configuring tools, suggested analyses).
C11 Improving understandability for non-experts (trustworthy presentation, representation per GP5, uncertainty warnings).

Epilogue / Glossary. Terminological clarification — "workflow mining", "automated business process discovery", "business process intelligence" — and a glossary.

Key claims

Process mining bridges data science and process science; it is the data-driven arm of BPM and uses real event data rather than modelled or assumed behaviour.
The three-type taxonomy (discovery / conformance / enhancement) is the definitional frame. Later refined into the cartography / auditing / navigation × offline / online 2×3 framework in Van der Aalst's 2016 textbook.
Event data quality governs achievable insight. The five-star scale (★–★★★★★) is a practical audit tool — reliable mining typically requires ★★★ or above.
Six guiding principles are non-negotiable for sound practice; they double as a review checklist.
Eleven open challenges (C1–C11) defined the research programme of the decade that followed the manifesto.
Operational support (detect / predict / recommend) is named as a critical research direction — the conceptual seed of the PPM and PrPM sub-fields.
Process mining is inherently a continuous activity aligned with the BPM lifecycle, not a one-off project.

Framing distinctions introduced

Process mining ≠ data mining. Data mining is process-agnostic; process mining treats concurrency, cases, and lifecycles as first-class.
Process mining ≠ control-flow discovery. Discovery is one of three types; organisational, case, and time perspectives are equally first-class.
Process mining ≠ offline. Online operational support is a valid (and important) mode.
Descriptive vs normative model. The same alignment produces different actions: descriptive — fix the model; normative — fix the process.
Lasagna vs spaghetti processes — though the terminology is richest in the textbook, the dichotomy is visible in the manifesto's C2 (diverse log characteristics) and underlies much of §4.

Textbook counterpart. sources/2011-vanderaalst-process-mining-book (1st ed. 2011, 2nd ed. 2016) — same author/community, same vocabulary, written concurrently. The manifesto is the rallying call; the textbook is the pedagogy.
Seed of PPM and PrPM. Challenge C8 (operational support) and the detect / predict / recommend trichotomy seed the entire PPM corpus (raw/Predictive process monitoring/). Every PPM method in this wiki extends the predict branch; every PrPM method (Kubrak et al. 2022) extends the recommend branch.
Successor manifesto. AI-Augmented BPMS (Dumas et al. 2023) reframes the same ambitions through AI augmentation; APM Manifesto (2026) reframes them through agent-centricity and explicitly identifies this manifesto as the exemplar that "stood the test of time as a rallying call for the community, facilitating the establishment of a data science tradition within BPM research."
Ongoing relevance of the challenges. Several challenges are still open in 2026 — notably C3 (benchmarks), C4 (concept drift), C7 (cross-organisational/privacy), C8 (runtime operational support beyond single cases), C10/C11 (usability and understandability for non-experts).
Authorial continuity. Many signatories recur as authors of later BPM works in this wiki: entities/wil-van-der-aalst, entities/marlon-dumas, entities/marcello-la-rosa, entities/fabrizio-maggi, entities/marco-montali, entities/jan-mendling, entities/barbara-weber, Boudewijn van Dongen, Arya Adriansyah.

Connections

Concepts: concepts/process-discovery · concepts/conformance-checking · concepts/operational-support · concepts/process-mining-spectrum · concepts/process-model-quality · concepts/lasagna-spaghetti-processes · concepts/predictive-process-monitoring · concepts/prescriptive-process-monitoring · concepts/concept-drift · concepts/bpm-lifecycle · concepts/business-process · concepts/process-architecture

Frameworks: frameworks/bpmn · frameworks/declare

Methods: methods/process-mining-basics · methods/process-discovery-methods · methods/agent-oriented-process-mining

Authors (entities): entities/wil-van-der-aalst · entities/marlon-dumas · entities/marcello-la-rosa · entities/fabrizio-maggi · entities/marco-montali · entities/jan-mendling · entities/barbara-weber

Related sources: sources/2011-vanderaalst-process-mining-book · sources/2018-dumas-fundamentals-of-bpm · sources/2023-dumas-ai-augmented-bpms · sources/2026-calvanese-agentic-bpm-manifesto · sources/2023-chapela-campa-augmented-process-execution · sources/2021-dumas-process-mining-2-from-insights-to-action