---
title: "Process Mining Manifesto"
type: source
tags: [process-mining, manifesto, bpm, foundational, ieee-task-force, guiding-principles]
authors: [van der Aalst Wil; Adriansyah Arya; de Medeiros Ana Karla Alves; Arcieri Franco; Baier Thomas; Blickle Tobias; Bose Jagadeesh Chandra; van den Brand Peter; Brandtjen Ronald; Buijs Joos; Burattin Andrea; Carmona Josep; Castellanos Malu; Claes Jan; Cook Jonathan; Costantini Nicola; Curbera Francisco; Damiani Ernesto; de Leoni Massimiliano; Delias Pavlos; van Dongen Boudewijn F.; Dumas Marlon; Dustdar Schahram; Fahland Dirk; Ferreira Diogo R.; Gaaloul Walid; van Geffen Frank; Goel Sukriti; Günther Christian W.; Guzzo Antonella; Harmon Paul; ter Hofstede Arthur; Hoogland John; Ingvaldsen Jon Espen; Kato Koki; Kuhn Rudolf; Kumar Akhil; La Rosa Marcello; Maggi Fabrizio M.; Malerba Donato; Mans Ronny S.; Manuel Alberto; McCreesh Martin; Mello Paola; Mendling Jan; Montali Marco; Motahari-Nezhad Hamid R.; zur Muehlen Michael; Munoz-Gama Jorge; Pontieri Luigi; Ribeiro Joel; Rozinat Anne; Seguel Pérez Hugo; Seguel Pérez Ricardo; Sepúlveda Marcos; Sinur Jim; Soffer Pnina; Song Minseok; Sperduti Alessandro; Stilo Giovanni; Stoel Casper; Swenson Keith D.; Talamo Maurizio; Tan Wei; Turner Chris; Vanthienen Jan; Varvaressos George; Verbeek Eric; Verdonk Marc; Vigo Roberto; Wang Jianmin; Weber Barbara; Weidlich Matthias; Weijters Ton; Wen Lijie; Westergaard Michael; Wynn Moe]
year: 2012
venue: "Business Process Management Workshops 2011, LNBIP 99, Springer: 169–194"
doi: "10.1007/978-3-642-28108-2_19"
kind: paper
raw_path: "raw/Process Frameworks & BPM/Process Mining Manifesto.pdf"
key_claims:
  - Process mining sits between computational intelligence / data mining on one side and process modelling / analysis on the other; it extracts knowledge from event logs rather than assumed processes.
  - Three basic types of process mining - discovery (log → model), conformance (log + model → diagnostics), and enhancement (log + model → improved model).
  - Process mining is not limited to control-flow; organisational, case, and time perspectives are equally first-class.
  - Process mining is not a special case of data mining and not restricted to offline analysis; it can support runtime operational support (detect, predict, recommend).
  - Six guiding principles govern good practice - event data as first-class citizens, question-driven log extraction, support for concurrency/choice/basic control-flow, events-to-model linkage, models as purposeful abstractions, process mining as a continuous process.
  - Five-level event-log maturity scale (★ to ★★★★★) relates log quality to the trustworthiness of mining results.
  - The L* life-cycle of a process mining project has five stages - plan and justify, extract, create control-flow model + connect log, create integrated model, provide operational support.
  - Eleven challenges (C1–C11) frame the research agenda - finding/merging/cleaning event data, coping with log diversity, benchmarks, concept drift, representational bias, balancing quality criteria, cross-organisational mining, operational support, combining with other analyses, usability and understandability for non-experts.
  - Event logs should be treated as first-class citizens, not as by-products of information systems.
created: 2026-04-13
updated: 2026-04-21
sources: []
---

# 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 [[concepts/process-mining-spectrum|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.
- **[[concepts/operational-support|Operational support]]** (detect / predict / recommend) is named as a critical research direction — the conceptual seed of the [[concepts/predictive-process-monitoring|PPM]] and [[concepts/prescriptive-process-monitoring|PrPM]] sub-fields.
- Process mining is inherently a **continuous** activity aligned with the [[concepts/bpm-lifecycle|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.

## Positioning vs related work in this wiki

- **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 [[concepts/predictive-process-monitoring|PPM]] corpus (`raw/Predictive process monitoring/`). Every PPM method in this wiki extends the *predict* branch; every PrPM method ([[sources/2022-kubrak-prescriptive-ppm-slr|Kubrak et al. 2022]]) extends the *recommend* branch.
- **Successor manifesto.** [[sources/2023-dumas-ai-augmented-bpms|AI-Augmented BPMS (Dumas et al. 2023)]] reframes the same ambitions through AI augmentation; [[sources/2026-calvanese-agentic-bpm-manifesto|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 ([[concepts/concept-drift|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]]