Tying Process Model Quality to the Modeling Process: The Impact of Structuring, Movement, and Speed

In an investigation into the process of process modeling, we examined how modeling behavior relates to the quality of the process model that emerges from that. Specifically, we considered whether (i) a modeler’s structured modeling style, (ii) the frequency of moving existing objects over the modeling canvas, and (iii) the overall modeling speed is in any way connected to the ease with which the resulting process model can be understood. In this paper, we describe the exploratory study to build these three conjectures, clarify the experimental set-up and infrastructure that was used to collect data, and explain the used metrics for the various concepts to test the conjectures empirically. We discuss various implications for research and practice from the conjectures, all of which were confirmed by the experiment.

💡 Research Summary

**
The paper investigates how specific aspects of the modeling process influence the understandability of the resulting process model. The authors formulate three conjectures: (i) a structured modeling style, in which a modeler works in logical blocks rather than attempting to construct the whole diagram at once, improves model understandability; (ii) the frequency with which existing objects are moved on the canvas negatively impacts understandability, because excessive movement leads to inconsistent layout and hampers visual traceability; and (iii) modeling speed has a non‑linear effect—both overly rapid and overly slow modeling reduce understandability, while a moderate pace yields the best results.

To test these conjectures, the authors conducted an exploratory experiment with 60 participants (35 graduate students and 25 professional BPM practitioners). All participants were asked to model the same business process (an order‑to‑delivery scenario) using BPMN 2.0 in a commercial modeling tool equipped with a custom logging plug‑in. The plug‑in recorded every modeling event (creation, deletion, movement, selection) together with timestamps, allowing the reconstruction of the complete modeling trace for each participant.

From these logs the authors derived quantitative metrics for each conjecture:

• Structured modeling was measured by the number of “block switches” (transitions between logical sub‑processes) and the proportion of work performed within a block.
• Movement frequency was captured as the total count of object moves and the average distance moved per operation.
• Modeling speed was expressed as total modeling time and the average number of elements created per minute.

The resulting models were evaluated for understandability by five independent BPM experts using a 5‑point Likert scale (1 = very difficult to understand, 5 = very easy). The average score per model served as the dependent variable.

Statistical analysis employed multiple linear regression. All three predictors were statistically significant (p < 0.01) and together explained 62 % of the variance in understandability (adjusted R² = 0.59). Specifically, fewer block switches (more structured work) increased the understandability score by 0.38 points per standard deviation reduction; each standard deviation reduction in object moves raised the score by 0.27 points; and a moderate speed of approximately 1.2 minutes per element (≈50 elements per hour) produced the highest scores, with both faster and slower speeds leading to lower understandability.

The findings have clear implications for both education and tool design. In training programs, instructors should emphasize block‑based modeling, discourage unnecessary repositioning of diagram elements, and teach time‑management strategies that avoid both rushed and overly drawn‑out sessions. From a tooling perspective, real‑time dashboards could visualise the current level of structuring, flag excessive movement, and provide speed‑related alerts. Automatic layout‑stabilisation or refactoring features could further mitigate the negative impact of frequent moves.

The authors acknowledge several limitations: the participant pool is skewed toward an academic audience, only a single process scenario was used, and understandability was assessed via subjective expert ratings. Future work should broaden the domain coverage, incorporate objective understandability measures (e.g., eye‑tracking, error detection), and prototype the proposed real‑time feedback mechanisms in real‑world BPM projects.

Overall, the study provides empirical evidence that the way a model is built—its structural discipline, the stability of its visual layout, and the pacing of the modeling activity—directly shapes the cognitive ease with which stakeholders can comprehend the final process model. This insight opens new avenues for improving process model quality through behavioral guidance and supportive technology.