Untidy Data: The Unreasonable Effectiveness of Tables

Area of Science:

• Human-Computer Interaction
• Data Visualization
• Information Science

Background:

• Data tables, particularly spreadsheets, are often viewed as a preliminary step in data analysis.
• However, they remain central to the information ecosystem for many 'data workers'—individuals who use data professionally but not as data scientists.
• Complex analytical tools can abstract data, limiting direct user interaction.

Purpose of the Study:

• To qualitatively study how data workers interact with and reason about data in table form.
• To understand the broader role of data tables in the sensemaking pipeline.
• To explore the potential of interactive tables as a visualization idiom.

Main Methods:

• Qualitative study involving data workers.
• Observation and analysis of user interaction with data tables.
• Analysis of how users reshape, augment, and reorganize data for sensemaking.

Main Results:

• Data tables serve a purpose beyond initial data cleanup, being used throughout the analytics process.
• Users actively "get their hands on" data, reorganizing, marking up, and layering information.
• Direct interaction with human-readable tables is cognitively important for understanding data meaning and utility.

Conclusions:

• Interactive data tables are a significant visualization idiom, not just a preparatory tool.
• The direct data interaction afforded by tables offers a valuable design space for visual analytics.
• Enhancing human-data interaction flexibility in visual analytics tools can enrich the sensemaking process.