Calvin Yau
Jul 22 · 4 min read

Improving time-series charts for communication.

Our work is motivated by the need to support data analysis for first responders. (Photo by Saffu on Unsplash.)

If a tree falls in the forest and no one is there to hear it, does it make a sound? Analogously, if an analyst writes a report about data and no one understands it, does it make a difference? In our new work, we try to improve the way analysts can communicate insights about data by proposing a new visualization that shows both a high-level overview as well as important details in a single interactive chart.

With more and more data about our world being collected through sensors, smartphone apps, and social media, people increasingly want to make data-driven decisions. For example, a chief of police may want to allocate police resources for different patrol routes based on past crime reports. A Coast Guard manager may want to decide which station to close in response to lost funds to minimize delay in response time based on historical incident records. The advertising department of a company may want to update its strategy based on the customer demographics and its survey results, and so on. But the sheer scale of it all is a double-edged sword: merely browsing the data in a spreadsheet is, in most cases, impractical and no longer sufficient for understanding. Data analysis tools such as R, Tableau, and SAS have been developed to help people better explore vast amounts of data and identify insights to support decision making using computations and visual representations. Unfortunately, not all of the knowledge gained using these tools are communicated effectively to the people that end up making the decisions.

In surveying six decision-makers in multiple first responder communities, ranging from local police departments to the U.S. Coast Guard, we found that they all experienced challenges when working with data analysts. While they all understood the goals, challenges, and what information is needed to make the right decisions, they were often not familiar with choosing the appropriate analysis or visual representation for different decisions. They were also not aware of how different visual representations of data could be misleading. Finally, they rarely had the time to interactively explore the data themselves using modern analysis tools. As a result, decision-makers rarely benefit from such tools directly. More commonly, the data is explored by professional analysts who produce a summarized, filtered, and possibly biased report for the decision-makers. By not being able to follow along and understand how the report was created, the decision-makers have no way to judge the reliability of the findings or draw their own conclusions. In other words, there is basically a gap in communication between analysts and decision-makers.

Our recent EuroVis 2019 paper presents a new family of data visualizations that addresses this gap by summarizing a dataset while including important insights and context. These summary visualizations combine three components: representative data from the full dataset, analytical highlights that directly support decisions, and a data envelope that provides context by summarizing the rest of the data. The idea is to provide a visualization that accommodates the decision-makers’ knowledge of what to look for, their lack of training in data analysis, and limited available time. For example, for line graph showing stock prices over time, the representative data could be the average change, the highlights could be points when a stock should be bought or sold, and the data envelope would show the minimum and maximum values for all stocks on the market.

An example of our summarized line graph displaying 8 airline stock prices.

Beyond the stock market, this kind of time-series data is important because it is central to many data-driven disciplines such as medicine, science, and engineering. To understand the effectiveness of our new summarized line graph, we conducted a user study comparing the performance between our design and four other time-series visualizations. The result shows our summarized line graph to give more accurate results, indicating the potential for better data communication to the decision-makers using our visual summary design.

What does this all mean? Clearly, an analysis report that decision-makers cannot understand is next to useless. The road to effective communication between analysts and stakeholders is a long one, but our work is one step closer towards this goal by providing a new visualization that strikes a balance between summary and details. To learn more about the design of the summarized line graph and the evaluation, please see our recent EuroVis 2019 paper:

Sparks of Innovation: Stories from the HCIL

Research at the Human-Computer Interaction Laboratory at University of Maryland

Calvin Yau

Written by

PhD student in Computer Engineering at Purdue University

Sparks of Innovation: Stories from the HCIL

Research at the Human-Computer Interaction Laboratory at University of Maryland

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade