Visual analytics, by its nature, helps users navigate large and complex datasets. This data can often be too large, containing many dimensions or data items, necessitating some type of summary visualization. We define summarization methods in broad terms, collecting techniques into four general methods: aggregation, projection, subsampling, and filtering.

In this paper, we take a guided approach to surveying the state of visualization, concentrating on how different choices in summarization methods can affect how the summaries are used in practice. Breaking a little from standard visualization surveys, we use existing visualization taxonomies to help define a codebook for categorizing existing visual analytic systems. Using the quantitative content analysis methodology (QCA), we identify correlations between the type of summarization employed, the purpose of the visualization, the viewer tasks supported by the visualization, and the data type.

The result of this process is a set of 16 themes, which collectively define a series of challenges in designing summary visualizations, and suggestions for ensuring the efficacy of new designs.

The results of the survey are available online using the SurVis system.

This work was presented at EuroVis 2018 in Brno, Czech Republic.