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):

Option to complete — every reachable state can reach a final state.
Proper completion — when the final state is reached, no orphan tokens remain (Dumas: each token in a different end event).
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 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.

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