---
title: 7PMG — Seven Process Modeling Guidelines
type: concept
tags: [bpm, modelling, quality, guidelines, pragmatic-quality, understandability, error-probability]
sources:
  - "[[sources/2010-mendling-reijers-vanderaalst-7pmg]]"
  - "[[sources/2018-dumas-fundamentals-of-bpm]]"
created: 2026-05-04
updated: 2026-05-04
---

# 7PMG — Seven Process Modeling Guidelines

Seven empirically-grounded, language-agnostic guidelines for producing process models that are **easier to understand** and **less error-prone**. Synthesised by Mendling, Reijers and van der Aalst (2010) from prior empirical studies on model size, complexity, structuredness and label ambiguity.

Each guideline is a **transformation that preserves behaviour** (modulo branching bisimulation): the resulting model expresses the same possible executions, just in a more readable form.

## The seven guidelines

| # | Guideline | Empirical basis |
|---|---|---|
| **G1** | Use as few elements as possible | Size correlates with both error probability and lower understandability (SAP Reference Model 600 EPCs; 2000 industrial EPCs) |
| **G2** | Minimise the routing paths per element | Connector degree (in-arcs + out-arcs) negatively correlates with understanding and positively with errors |
| **G3** | Use one start and one end event | Multiple starts/ends correlate with higher error probability; workflow engines often require single start/end; a single start/end is required to apply [[concepts/soundness]] checks |
| **G4** | Model as structured as possible | Each split connector should be matched by a join of the same type ("balanced brackets"). Unstructured models are both more error-prone and harder to understand |
| **G5** | Avoid OR routing elements | OR-join semantics carries known paradoxes and implementation problems; AND/XOR-only models are less error-prone empirically |
| **G6** | Use verb-object activity labels (e.g. "Inform complainant") | Significantly less ambiguous and more useful than action-noun ("Complaint analysis") or other styles in controlled n=29 experiment |
| **G7** | Decompose model if it has more than 50 elements | Error probability rises above 50% beyond this threshold; split single-entry/single-exit subcomponents into sub-process models |

## Empirical ground truth

7PMG distinguishes itself from prior frameworks (SEQUAL, GoM) by being **operational** and **empirically grounded**. The three measurement axes:

- **Process model understanding** — questionnaire studies (n=73) measuring comprehension against structural metrics.
- **Error probability** — relaxed soundness (600 SAP EPCs) and EPC Soundness (2000 industrial EPCs); structural-metric-based prediction functions for error probability.
- **Label ambiguity** — controlled labelling-style experiments (n=29 postgraduates).

This grounding lets each guideline cite a specific study; cf. [[sources/2010-mendling-reijers-vanderaalst-7pmg]] §3.2.

## Guideline interactions

The guidelines can **conflict**: applying **G2** (reduce connector degree) typically requires *adding* connectors, increasing element count and tensing with **G1**. The 7PMG paper offers a workshop-derived priority ordering from 21 professional modellers (German + Dutch practice) but acknowledges that a theoretically-motivated, empirically-validated prioritisation is out of scope.

A practical heuristic: apply guidelines in the order they reduce model size first (G1, G7), then improve structure (G3, G4), then handle local routing (G2, G5), then labels (G6).

## Scope and limitations

- **Control-flow only**. 7PMG does not cover data, organisation, compliance, or non-functional dimensions.
- **Behaviour-preserving** transformations only — 7PMG describes transformations within an equivalence class of models with the same behaviour. *Redesign* (changing what the process does) is a separate activity (see [[methods/process-redesign-heuristics]], [[concepts/devils-quadrangle]]).
- **Heuristic, not formal**. G1, G2, G7 are size/complexity heuristics. They reduce error probability but do not guarantee correctness — for that, pair with [[concepts/soundness]] verification.
- **Empirical foundation is EPC-heavy**. Studies were conducted predominantly on EPCs and SAP-style models. Direct re-validation on BPMN corpora has been partial (replication studies exist but are smaller).

## Relationship to other quality frameworks

| Framework | Layer | Relation to 7PMG |
|---|---|---|
| [[concepts/sequal-framework]] (Krogstie) | Theoretical (semiotic) | 7PMG operationalises SEQUAL's pragmatic + (partially) syntactic quality |
| [[concepts/process-model-quality]] (Dumas) | Textbook tripartite | 7PMG is the operational toolkit for pragmatic quality (§5.4.4) |
| [[concepts/soundness]] (van der Aalst) | Formal correctness | Complementary — 7PMG addresses understandability/error-prevention; soundness addresses structural correctness |
| [[concepts/process-model-complexity-metrics]] | Bottom-up metrics | 7PMG distils these metrics into actionable rules |

## Application — worked example

The 7PMG paper transforms a 37-element complaint-handling EPC (multiple starts/ends, OR-join, mixed labels) into a 31-element model with single start/single end, structured nesting, no OR-join, verb-object labels — same behaviour, demonstrably more readable. See [[sources/2010-mendling-reijers-vanderaalst-7pmg]] §3.4 for full walk-through.

## Practical use as a checklist

For an as-is BPMN model produced from interviews ([[syntheses/interview-structuring-for-process-models]]):

1. **Count elements** (G1) — flag if growing.
2. **Single start, single end?** (G3) — required.
3. **Unbalanced splits/joins?** (G4) — restructure where possible.
4. **OR-gateway present?** (G5) — replace with XOR + default branch unless inclusive semantics is genuinely needed.
5. **Connector degree ≥ 4?** (G2) — split into smaller connectors.
6. **All labels verb-object?** (G6) — re-label.
7. **Total elements > 50?** (G7) — extract sub-processes.

## Related
[[concepts/process-model-quality]] · [[concepts/sequal-framework]] · [[concepts/soundness]] · [[concepts/process-model-complexity-metrics]] · [[frameworks/bpmn]] · [[entities/jan-mendling]] · [[entities/hajo-reijers]] · [[entities/wil-van-der-aalst]]

## Source
[[sources/2010-mendling-reijers-vanderaalst-7pmg]] (primary) · [[sources/2018-dumas-fundamentals-of-bpm]] (textbook citation in §5.4.4)