An Empirical Study of Task-Specific Limitations of the Overview+Detail Technique for Interactive Time Series Analysis

Interactive visualisation and analysis of time series data is a critical part of many data-driven optimisation processes, particularly in Industry 4.0 and Smart Manufacturing. Time series visualisation enables data analysts or domain experts to visually identify problems such as missing values, sensor drift, precision degradation, or faulty data, before or after algorithmic analysis. A common technique to support the visual exploration of large time series is the overview+detail (O+D) technique. O+D provides both detail and context information by displaying a detailed view showing the actual data and a thumbnail overview for showing its context. User studies have shown that users analyse and navigate data sets more efficently and effectively with than without O+D, but that this strongly depends on the task and the nature and amount of the data to be displayed. We present results of a quantitative user study that was performed on Amazon Mechanical Turk with 95 participants to identify scenarios in which O+D could not effectively solve the challenge of visualising large time series. By this, we identify potential usability issues of O+D for typical time series analysis tasks and discuss their origins. For each of these usability issues, we also propose alternative interaction and visualisation designs or other strategies to maintain good usability, even for challenging task types and data densities.