DynDash: Multiple Coordinated Dashboards for Exploratory Data Analysis

Exploratory visual analysis is a good way to find novel, unforeseen insights in large amounts of data. Most existing Visual Analytics applications require the user to manually specify charts first. Such a depth-first strategy is unsuitable for novices due to lack of expertise. Voyager 2 proposed a mixed-initiative approach that blends manual specification with automatic chart recommendations. To facilitate iterative sensemaking, we extend this approach through enabling users to build multiple dashboards, each of which holds coordinated, fully interactive charts in a flexible layout. Together, this permits users to quickly get an overview of the data, while still being able to analyze details. We also present a new, non-intrusive filtering mechanism that enables creating, copying and editing of filters. We performed a qualitative study with novices and identified common usage patterns, which inform the design of future multi-dashboard Visual Analytics systems.