Showing Flow: Comparing Usability of Chord and Sankey Diagrams

Title of the Paper

Showing Flow: Comparing Usability of Chord and Sankey Diagrams

Paper Information

• Domain: Information Visualization and Human-Computer Interaction
• Keywords: Chord diagram, Sankey diagram, data flow visualization, radial layout, linear layout, user experience, graphical design

Research Background and Problem Statement

• Identified Problems or Challenges:

  • Chord diagrams and Sankey diagrams are common visualization techniques for representing data flows, but the specific impacts of these two layouts on user experience and information interpretation have not been thoroughly compared.
  • Radial layouts (Chord diagrams) face issues such as rotating entities and dense central links, which may reduce usability and evaluation efficiency.
  • Current understanding of the performance and usability of these layouts is limited, especially in specific application scenarios for flow data visualization.

• Why It Matters:

  • Information visualization design directly affects users' efficiency, accuracy, and satisfaction in analyzing data.
  • With the increasing complexity of datasets, selecting appropriate visualization layouts is crucial for supporting user understanding and decision-making.

• Research Motivation and Related Work:

  • Radial layouts are widely used across various fields (e.g., bioinformatics, engineering, medicine), but existing studies suggest that linear layouts (e.g., Sankey diagrams) may perform better in certain cases.
  • Related work has explored the advantages and disadvantages of radial and non-radial layouts, but there is a lack of specific design guidelines for visualizing flow data.

Proposed Solution

• Proposed Solution or Method:

  • Conducted a crowdsourced experiment comparing Chord diagrams and Sankey diagrams across five tasks (link existence checking, element search, value comparison, maximum/minimum value identification, link counting).
  • Metrics include completion time, error rate, subjective effort perception (NASA-TLX scale), and overall user preference.

• Innovative Aspects:

  • Systematically compared Chord and Sankey diagrams in terms of user experience and performance in flow data visualization, providing design optimization suggestions.
  • Verified the potential limitations of radial layouts for user usability and summarized the advantages of linear layouts.

• Implementation Steps and Techniques:

  • Used a custom JavaScript library based on D3.js to create equivalent Chord and Sankey diagrams, ensuring fair graphical design.
  • Conducted experiments across four datasets (debt transfer, migration flow, mobile brand switching, spatial industry investment) and five question types.
  • Measured participants' task completion time and error rate while collecting subjective feedback and preferences.

Research Findings

• Specific Results:

  • The average task completion time for Chord diagrams was 3.7 seconds longer than for Sankey diagrams, with a more significant gap in the first round of tasks (9 seconds longer).
  • Chord diagrams had a higher error rate (on average, 0.1 more errors), with a more pronounced difference in the first question round (0.42 more errors).
  • Using the NASA-TLX scale, participants perceived Chord diagrams as requiring higher mental effort, being more frustrating, and overall performing worse than Sankey diagrams.
  • In the preference survey, 35 out of 42 participants found Sankey diagrams faster, 43 found them more accurate, and 42 preferred them overall.

• Advantages Compared to Existing Solutions:

  • Quantified the performance differences between the two layouts in flow data visualization, providing new design considerations.
  • Specifically highlighted the limitations of Chord diagrams in initial usage scenarios, aiding designers in optimizing visual interaction.

• Experimental or Evaluation Results:

  • Consistent results across four different datasets, demonstrating the overall superior performance and user preference for Sankey diagrams.
  • Moderate data size (8-15 nodes, 50-100 links), covering common visualization task scenarios.

• Limitations and Future Directions:

  • Chord nodes were not split into separate input and output links; future studies could test their impact on performance.
  • Data sets were of moderate size; further research is needed on visualization effects for larger-scale data.
  • Focused on directed link datasets; future work could extend to undirected link scenarios.
  • Limited task scope; more complex tasks for analyzing flow data should be included.
  • Did not assess long-term performance changes or potential improvements after users gain expertise.
  • Future research should explore interactive chart designs to investigate the impact of dynamic interactions on performance.

Conclusion

This study clearly demonstrated the advantages of Sankey diagrams in flow data visualization, providing practical guidance for graphic designers and visualization tool developers. While Chord diagrams retain aesthetic value in certain contexts, from a user experience perspective—particularly in initial usage scenarios—the linear layout of Sankey diagrams is better suited to support users in quickly analyzing and understanding data flows.