Interver: Drilling into Categorical-Numerical Relationships

Abstract

Data analytics is increasingly performed by non-expert analysts (e.g., casual business users). In this context, future analytics tools need easy-to-use techniques to reveal relations between columns of data in a spreadsheet or table. For example, a market analyst, may want to find if industry categories and funding amounts are related: i.e., if some industries receive amounts within distinctive intervals. Traditional filtering and script-based querying poorly support non-expert users in such explorations because they require iterative parameter adjusting and query writing until a meaningful result is found. In this paper, we focus on supporting the analysis of relationships between categorical and numerical columns. We present a novel visualization, Interver, which dynamically reveals insights as the user selects an interval within the relationship. With a concrete scenario, specific analysis tasks, and an informal evaluation, we show how Interver can help non-expert analysts self-serve and answer realistic questions.