Charts exist to inform decisions, not decorate slides. Good visualizations match the message, respect perceptual psychology, and stay honest about uncertainty.
Choose the right chart
- Comparison — bar charts (sorted), dot plots
- Trend over time — line charts with clear time axis
- Distribution — histogram, box plot (local matplotlib/seaborn)
- Relationship — scatter plot with transparency for overlap
- Part-to-whole — stacked bars sparingly; avoid pie charts with many slices
Design habits
- Label axes with units; title states the takeaway
- Start bar chart baselines at zero; truncate line axes only when intentional
- Use color for meaning, not decoration—check color-blind palettes
- Remove chartjunk (3D, heavy grids, dual y-axes unless expert)
Honesty
Show sample size, confidence intervals, or ranges when stakes are high. Anomalies and missing data belong in footnotes.
Tooling path
Playground uses stdlib summaries; locally use matplotlib and seaborn with Pandas DataFrames. Dashboards may use BI tools—principles stay the same.
Important interview questions and answers
- Q: Why avoid pie with many slices?
A: Humans compare angles poorly—bars compare lengths more accurately. - Q: Truncated y-axis?
A: Can exaggerate differences—use only when audience understands scale.
Self-check
- What chart type shows change over time?
- Why label axes with units?
- When is a bar chart baseline at zero required?
Tip: Label axes and units—never ship unlabeled charts.
Interview prep
- Chartjunk?
Decoration that obscures data—avoid.