Towards Visualization Thumbnail Designs That Entice Reading Data-Driven Articles

Hwiyeon Kim, Joohee Kim, Yunha Han, Hwajung Hong, Oh-Sang Kwon, Young-Woo Park, Niklas Elmqvist, Sungahn Ko, Bum Chul Kwon

TL;DR

This paper explores how to design visualization thumbnails for data journalism that are both "inviting" (attracting clicks) and "interpretable" (conveying the main point). Through a survey of industry practices and a user study, the researchers found that readers significantly prefer thumbnails that use visual highlights and human-recognizable icons over simple resized charts or generic photographs.

What Problem Does This Paper Address?

In the era of data journalism, news organizations (like The New York Times or FiveThirtyEight) increasingly rely on visualizations to tell complex stories. However, the "first impression" of these stories often comes from a tiny thumbnail image on a social media feed or a homepage.

The challenge is a classic trade-off: a thumbnail must be small enough to fit a layout but detailed enough to be useful. Simple resized charts are often too cluttered and "cold" to entice readers, while generic photos might attract clicks but fail to communicate the data's actual message. This paper seeks to define a design space for "visualization thumbnails" that balances these competing needs for appeal and informativeness.

Methods

The authors employed a three-phase approach:

  1. Survey of Practice: They collected and codified 67 visualization thumbnails from top-tier news outlets, identifying 14 basic and 4 additional component types (like axes, data labels, and "Human Recognizable Objects" or HROs).
  2. Expert Interviews: They interviewed six industry practitioners (data journalists and designers) to understand the "why" behind current designs, uncovering goals like brand reinforcement and aesthetic appeal.
  3. User Study ($n=161$): Using the insights from the first two phases, they designed 16 thumbnails across four categories:
  4. Resized: Direct, smaller versions of the original chart.
  5. GNRD (Graphics Not Related to Data): Charts paired with a relevant photograph.
  6. Highlight: Simplified charts with red blocks or data labels emphasizing a specific point.
  7. HRO (Human Recognizable Objects): Charts using icons or logos (e.g., an Apple logo or handcuffs) as visual legends.

To analyze preferences, they used a Thurstonian model and Monte Carlo Markov-Chain (MCMC) sampling to estimate the probability of one design being preferred over another.

Datasets

The researchers did not use a standard machine learning dataset. Instead, they created:

  • A Corpus of 139 Data-Rich Articles: Collected from outlets including The Economist, FiveThirtyEight, and The Wall Street Journal (Nov–Dec 2018).
  • Experimental Stimuli: 16 custom-designed thumbnails based on four real-world news topics (LeBron James' influence, Trump's approval ratings, Apple vs. Microsoft stocks, and the Kavanaugh confirmation).

Key Insights and Ideas

  • The "Visual Legend" Effect: HROs (icons/logos) act as intuitive legends. Readers can instantly identify what the data represents without squinting at tiny text, making the chart feel more accessible.
  • The "Visual Guide" Strategy: Highlights (like red blocks over a specific time range) act as anchors. They tell the reader exactly where to look and what the "takeaway" is, which increases the perceived informativeness of the thumbnail.
  • The Axis Paradox: While detailed axes and tick marks increase a reader's trust and the perceived "professionalism" of an article, readers rarely actually use them to interpret the data in a thumbnail. In fact, full axes often lead to "clutter," which can deter clicks.
  • Double-Edged Photographs: While practitioners thought photographs (GNRD) would be the most inviting, the user study showed they were less preferred than data-centric designs. Photos can be "inspiring" but often lead to open, and sometimes incorrect, interpretations of the data.

Key Results

  • Preference Ranking: The Highlight and HRO designs were the clear winners. Statistical analysis showed a 99.5% probability that a Highlight design would be preferred over a GNRD (photo-based) design.
  • Interpretability vs. Invitation: "Highlight" designs were rated highest for being informative (27% of comments), while "HRO" designs were praised for being eye-catching (26% of comments).
  • Resized Charts Lag Behind: Despite being common in industry, "Resized" thumbnails were the least preferred, often described by participants as "unfriendly" or "hard to read."

Limitations and Future Work

  • Chart Type Restriction: The user study focused exclusively on line charts. The authors acknowledge that the roles of components might shift for bar charts, maps, or scatterplots.
  • Single Size: The study used a fixed thumbnail size (343px width). Future work could investigate how "micro-thumbnails" (e.g., on smartwatches) require even more aggressive simplification.
  • Static vs. Dynamic: The study only looked at static images. As news platforms evolve, the role of animated or interactive thumbnails remains an open research question.
  • Message vs. Summary: The authors suggest further research into whether a thumbnail should convey a single "fact" or attempt to summarize the entire narrative arc of a story.

Summary updated

Abstract

As online news increasingly include data journalism, there is a corresponding increase in the incorporation of visualization in article thumbnail images. However, little research exists on the design rationale for visualization thumbnails, such as resizing, cropping, simplifying, and embellishing charts that appear within the body of the associated article. Therefore, in this paper we aim to understand these design choices and determine what makes a visualization thumbnail inviting and interpretable. To this end, we first survey visualization thumbnails collected online and discuss visualization thumbnail practices with data journalists and news graphics designers. Based on the survey and discussion results, we then define a design space for visualization thumbnails and conduct a user study with four types of visualization thumbnails derived from the design space. The study results indicate that different chart components play different roles in attracting reader attention and enhancing reader understandability of the visualization thumbnails. We also find various thumbnail design strategies for effectively combining the charts’ components, such as a data summary with highlights and data labels, and a visual legend with text labels and Human Recognizable Objects (HROs), into thumbnails. Ultimately, we distill our findings into design implications that allow effective visualization thumbnail designs for data-rich news articles. Our work can thus be seen as a first step toward providing structured guidance on how to design compelling thumbnails for data stories.
BibTeX
@article{kim2024towards,
  author = {Hwiyeon Kim and Joohee Kim and Yunha Han and Hwajung Hong and Oh-Sang Kwon and Young-Woo Park and Niklas Elmqvist and Sungahn Ko and Bum Chul Kwon},
  title = {{Towards Visualization Thumbnail Designs That Entice Reading Data-Driven Articles}},
  journal = {IEEE Trans. Vis. Comput. Graph.},
  year = {2024},
  url = {https://openreview.net/forum?id=uf6r4WQKVw},
}

Loading related content...

Figure