Tables of numbers are incomprehensible at scale — humans cannot intuitively grasp patterns, trends, or outliers from raw data.
Representing data graphically to enable pattern recognition, comparison, and communication of insights.
- Select chart type based on data and message (bar, line, scatter, heatmap)
- Map data variables to visual channels (x, y, color, size, shape)
- Apply design principles: minimize clutter, highlight key findings
- Iterate: refine based on what the audience needs to understand
- Communicate: use visualizations to tell a data-driven story
- Channel-effectiveness: some visual channels (position) are better than others (color)
- Audience-dependent: different charts for technical vs non-technical audiences
- Purpose-driven: exploratory (many charts, quick) vs explanatory (few charts, polished)
- Truth-telling: visualizations must accurately represent the underlying data
- Built from: EDA — visualization is EDA’s primary tool
- Builds into: Data Science — communication is the final step
- Related: Feature Engineering — visualizations reveal feature relationships
- Related: Data Wrangling — visualize data quality issues
- Misleading axes (non-zero baselines, truncated ranges) distort perception
- Over-plotting in scatter plots hides density and patterns
- Color blindness: red-green comparisons are inaccessible
- Chart junk: unnecessary decorations distract from the data