Data visualization turns numbers into pictures so that people can see patterns, comparisons and outliers that a table hides. A good chart answers a specific question honestly and quickly; a bad one decorates, confuses or misleads. It rests on how human vision works as much as on any software.
Key concepts
- Choose the chart for the question: lines for change over time, bars for comparing amounts, scatter plots for relationships, histograms for distributions.
- Visual encoding: position and length are read accurately; area, angle and colour intensity much less so.
- Honest scales: bars start at zero, axes are labelled with units, and truncated axes are marked.
- Reduce clutter: remove anything that does not carry information, and annotate what matters.
- Colour and access: use colour sparingly, check it works for colour-blind readers and never rely on it alone.
Vocabulary
- Encoding: the mapping from a data value to a visual property such as position or colour.
- Small multiples: a grid of similar small charts that makes comparison easy.
- Overplotting: so many points that they hide each other.
- Dashboard: a screen of charts for monitoring, as opposed to one chart that tells a story.
What to study, in order
- Plot one dataset as a table, then as a line, a bar and a scatter plot, and compare what each shows.
- Learn a plotting library in a language you know and reproduce a chart from a published article.
- Study perception: why position beats area, and what makes a colour scale readable.
- Redesign a bad chart: fix its scale, remove clutter and add a clear title that states the finding.
- Build an interactive view with filters and tooltips, and test it with someone new to the data.
- Learn to handle large data: aggregation, binning and sampling.