[Paper] Tables or Sankey Diagrams? Investigating User Interaction with Different Representations of Simulation Parameters

Source: arXiv - 2601.10232v1

Overview

The paper investigates whether interactive Sankey diagrams can help engineers understand complex parameter dependencies better than the classic spreadsheet‑style tables that dominate many configuration tools. By running a structured heuristic evaluation, the authors show that visualizing the flow of parameters dramatically cuts cognitive load and the number of steps needed to complete typical engineering tasks.

Key Contributions

• Empirical evidence that flow‑based visualizations (Sankey diagrams) halve the perceived effort (‑51 % PURE score) compared to tabular interfaces.
• Interaction‑cost reduction: users performed 56 % fewer interaction steps to achieve the same goals.
• Design guideline: a concrete prototype that makes parameter relationships explicit, eliminating the need for mental reconstruction.
• Cross‑domain relevance: although evaluated on a Computer‑Aided Engineering (CAE) tool, the findings apply to any configuration‑heavy software (cloud orchestration, database tuning, CI/CD pipelines, etc.).
• Methodological contribution: adaption of the PURE (Usability Evaluation) heuristic framework for assessing visualizations of software configuration data.

Methodology

1. Prototype Development – The researchers built an interactive Sankey‑based UI that maps input parameters to downstream simulation outputs, using animated flows to indicate magnitude and direction.
2. Comparison Baseline – A conventional spreadsheet view (rows = parameters, columns = values) served as the control condition.
3. PURE Heuristic Evaluation – Three domain experts (UX designer, simulation engineer, software developer) inspected both interfaces using the PURE checklist (usability, learnability, efficiency, error prevention, satisfaction).
4. Task Scenarios – Participants performed typical engineering tasks such as “identify which input influences a given output” and “modify a set of inter‑dependent parameters.”
5. Metrics Collected – PURE scores (lower = better) and the count of interaction steps (clicks, drags, selections) required to finish each task.

The approach stays lightweight—no large‑scale user study—yet provides a rigorous, expert‑driven comparison that is easy to replicate in other domains.

Results & Findings

What it means:

• The Sankey representation makes the structure of dependencies visible at a glance, so users spend less mental effort piecing together relationships.
• Fewer clicks and shorter task times translate directly into productivity gains, especially in environments where engineers repeatedly tweak large parameter sets.
• Higher confidence scores suggest that users trust the system more when they can see the “flow” of influence.

Practical Implications

• Tool Builders: Embedding Sankey‑style visualizations into configuration panels (e.g., Terraform UI, Kubernetes dashboards, DB tuning consoles) can reduce onboarding time and error rates.
• DevOps & Cloud Engineers: Quick identification of cascading effects when changing a resource limit or scaling rule becomes possible without digging through dozens of YAML entries.
• CAE & Simulation Software Vendors: Adding an interactive flow view can differentiate products and justify premium pricing by promising faster design iterations.
• Education & Training: Sankey diagrams serve as a pedagogical aid for teaching complex system interactions, making abstract dependencies concrete.
• Automation: The explicit graph of parameter relationships can be exported to downstream tools (e.g., impact analysis scripts, automated test generation) because the visual model already encodes the dependency graph.

Limitations & Future Work

• Expert‑only evaluation – The study relied on three specialists; broader user studies (including novices) are needed to confirm generalizability.
• Domain specificity – The prototype was tuned for CAE; adapting the visual encoding (e.g., handling thousands of nodes) may require additional scalability research.
• Static vs. Dynamic data – The current Sankey view reflects a snapshot of parameter values; future work could explore real‑time updates as simulations run.
• Integration challenges – Embedding Sankey diagrams into existing legacy tools may involve non‑trivial UI redesign and performance considerations.

The authors suggest extending the approach to other configuration‑intensive domains, investigating hybrid views (tables + Sankey), and conducting longitudinal field studies to measure long‑term productivity impact.

Authors

• Choro Ulan uulu
• Mikhail Kulyabin
• Katharina M Zeiner
• Jan Joosten
• Nuno Miguel Martins Pacheco
• Filippos Petridis
• Rebecca Johnson
• Jan Bosch
• Helena Holmström Olsson

Paper Information

• arXiv ID: 2601.10232v1
• Categories: cs.HC, cs.SE
• Published: January 15, 2026
• PDF: Download PDF