1 min read· by Awab Tech Lover
Data viz that doesn't lie
A short field-guide to picking chart types, sane axes, and colour palettes that respect your readers.
Data viz that doesn't lie
A good chart removes work from the reader's brain. A bad chart adds work — or worse, misleads.
Three rules I try to follow
1. Match the chart to the question
- Comparing categories? Bar chart.
- Showing change over time? Line chart.
- Part of a whole? Stacked bar (almost never a pie).
2. Start the y-axis at zero (usually)
Truncated axes exaggerate differences. If you must truncate, label it loudly.
3. Use colour intentionally
A palette of 12 colours is a palette of zero. Pick one accent, one neutral, and one "danger" colour, and let the data speak.
Tools I reach for
- Plotly for interactivity
- Matplotlib for print-quality output
- D3 when I need full control on the web
If your dashboard needs a legend with 15 entries, the dashboard is wrong, not the user.