---
title: Process Model Quality
type: concept
tags: [bpm, modelling, quality, validation]
sources:
  - "[[sources/2018-dumas-fundamentals-of-bpm]]"
  - "[[sources/2012-ottensooser-graphical-vs-textual]]"
  - "[[sources/2006-krogstie-sindre-jorgensen-revised-sequal-framework]]"
  - "[[sources/2010-mendling-reijers-vanderaalst-7pmg]]"
  - "[[sources/1998-vanderaalst-verification-of-workflow-nets]]"
created: 2026-04-13
updated: 2026-05-04
---

# Process Model Quality

Dumas et al. ([[sources/2018-dumas-fundamentals-of-bpm]], §5.4) decompose model quality into three orthogonal dimensions — a textbook-grade distillation of the **SEQUAL framework** ([[concepts/sequal-framework]], [[sources/2006-krogstie-sindre-jorgensen-revised-sequal-framework]]) which uses an eight-level semiotic structure to define quality theoretically. SEQUAL is the theoretical scaffolding; Dumas's tripartite is its actionable subset.

## Three dimensions

### Syntactic quality
Does the model respect the notation's rules? Verified mechanically: no dangling edges, gateway/join correctness, type constraints. Tool-enforceable.

### Semantic quality
Does the model correctly reflect the real process? Validated with domain experts. Requires:
- **Completeness** — all relevant behaviour captured.
- **Correctness** — no spurious behaviour represented.

### Pragmatic quality
Is the model understandable and usable for its audience? Governed by:
- Layout
- Label clarity
- Size / complexity
- Modelling guidelines (7PMG — Seven Process Modeling Guidelines)

## Modelling guidelines (§5.4.4) — 7PMG
The textbook bullets are the seven guidelines from [[sources/2010-mendling-reijers-vanderaalst-7pmg]] (see [[concepts/7pmg]] for full rationale and empirical basis):

- **G1**: Use as few elements as possible.
- **G2**: Minimise routing paths per element.
- **G3**: Use one start and one end event.
- **G4**: Model as structured as possible.
- **G5**: Avoid OR-joins.
- **G6**: Use verb-object activity labels.
- **G7**: Decompose large models (≥50 nodes).

## Syntactic quality — two rule classes

Per [[sources/2018-dumas-fundamentals-of-bpm]] §5.4.1, syntactic quality decomposes into:

### Structural rules (graph-level, mechanical)

**Element-level** (BPMN):
- **Activity**: ≥1 incoming AND ≥1 outgoing sequence flow.
- **Start event**: no incoming sequence flow.
- **End event**: no outgoing sequence flow.
- **Intermediate event**: ≥1 incoming AND ≥1 outgoing sequence flow.
- **Boundary event**: only intermediate catching events on activity borders.
- **Split gateway**: exactly 1 incoming, ≥2 outgoing.
- **Join gateway**: ≥2 incoming, exactly 1 outgoing.
- **(X)OR-split outgoing arcs** must bear conditions.
- **Sequence flow**: connects two flow nodes of the same pool; cannot cross pools.
- **Message flow**: connects activity/throwing-message-event in one pool to activity/catching-message-event in another pool.
- **Data association** (directed): data object ↔ activity / throwing-message-event; data store ↔ activity.

**Model-level**: all flow nodes must be on a path from start to end (connectedness).

### Behavioural rules — soundness

The formal correctness criterion ([[concepts/soundness]], [[sources/1998-vanderaalst-verification-of-workflow-nets]]):

1. **Option to complete** — every reachable state can reach a final state.
2. **Proper completion** — when the final state is reached, no orphan tokens remain (Dumas: each token in a *different* end event).
3. **No dead transitions** — every modelled activity is executable.

Behavioural rules forbid four anomalies:

| Anomaly | Concept page | Violates |
|---|---|---|
| Token stuck — instance can't progress | [[concepts/deadlock]] | (1) option to complete |
| Token cycles forever in loop | [[concepts/livelock]] | (1) option to complete |
| Multiple tokens on same flow | [[concepts/lack-of-synchronization]] | (2) proper completion |
| Activity never executable | [[concepts/dead-activity]] | (3) no dead activities |

Soundness is decidable in polynomial time for free-choice [[concepts/workflow-net|workflow nets]] — the practical class. Block-structured models ([[concepts/block-structure]]) are sound by construction.

## Bottom-up complexity metrics
[[concepts/process-model-complexity-metrics]] — size, density, average/max connector degree, CFC, mismatch — predict error probability and understandability. Cited as the empirical basis for 7PMG guidelines G1, G2, G4, G7.

## Empirical evidence on pragmatic quality

Ottensooser et al. ([[sources/2012-ottensooser-graphical-vs-textual]]) ran a within-subjects experiment with 196 participants across three universities, comparing BPMN diagrams to Cockburn-style written use cases:

- **Untrained readers ("business users") gain no significant understanding from BPMN alone** (Wilcoxon p=0.15); written use cases helped both trained and untrained readers (p<0.01 for both).
- **Joint presentation in the order UC → BPMN yielded maximal scores for all reader groups** (H5 supported at p<0.01 for BA and BU proxies); the reverse order (BPMN → UC) gave no significant lift over BPMN alone.
- Familiarity with flowcharts/BPMN predicts preference for graphical notation — graphical notations are a *learned* skill, not intuitively accessible.

Implication for pragmatic quality: a BPMN model scores poorly on pragmatic quality *for untrained audiences* regardless of 7PMG compliance. For mixed audiences, pair the diagram with a structured natural-language description presented first.

## Related
[[concepts/process-discovery]] · [[methods/process-discovery-methods]] · [[frameworks/bpmn]] · [[concepts/soundness]] · [[concepts/7pmg]] · [[concepts/sequal-framework]] · [[concepts/process-model-complexity-metrics]] · [[concepts/workflow-net]] · [[concepts/conformance-checking]] · [[syntheses/process-model-quality-and-soundness-evaluation-guide]]