20 Proven Rules to Create Powerful Data Visualizations

Data visualization is a common way to present data; a good chart can convey information efficiently, while a poor one confuses the audience.

About half of the human brain is dedicated to processing visual information, so well‑designed visualizations have a strong impact.

Following a few design rules and some aesthetic training enables anyone to create professional‑looking charts.

Good visualizations often follow proven design principles, as illustrated by the New York Times examples.

1. Choose the appropriate chart type

Using the wrong chart type can cause confusion. Select the type based on what you want to show; the article lists several common chart types.

2. Choose drawing direction based on positive/negative values

When drawing horizontal bar charts, place negative values on the left Y‑axis and positive values on the right; do not mix them on the same side. The same applies to vertical column charts.

3. Start bar charts from a zero baseline

Truncating the Y‑axis can distort perception. Starting from zero ensures users see an accurate representation of differences.

4. Use adaptive Y‑axis scaling for line charts

Always starting the Y‑axis at zero can flatten line charts. Adjust the scale to the data range so the line occupies roughly two‑thirds of the axis height.

5. Use line charts sparingly when time points are sparse

Line charts work well for dense time series (e.g., minute‑level stock data). When points are few and far apart, bar charts are clearer.

6. Avoid overly smooth line charts

Smooth lines may look appealing but can misrepresent the underlying data and hide actual data points.

7. Avoid dual‑axis charts

Dual‑axis charts save space but are hard to read and can mislead viewers into thinking the two series are directly comparable.

8. Limit the number of slices in a pie chart

Pie charts should contain no more than 5‑7 slices; extra small segments can be grouped into an “Other” slice.

9. Annotate directly on the chart

Charts without proper labels are meaningless. Direct labeling on the chart helps viewers understand data instantly.

10. Do not place values directly on pie slices

Putting values on top of slices reduces readability, especially for thin slices. Use external labels with clear pointers.

11. Sort pie‑chart slices

Unordered slices are hard to interpret. Place the largest slice at the 12‑o’clock position and arrange the rest clockwise in descending order.

12. Avoid random ordering in bar charts

Sort bars consistently—largest values at the top (horizontal) or left (vertical)—to reduce eye movement and improve readability.

13. Set appropriate width for donut charts

Donut charts free space for additional information, but if the ring is too narrow the chart becomes unreadable.

14. Keep styles simple; let data speak

Unnecessary styles (3D, shadows, gradients, excessive gridlines, decorative fonts) distract and can mislead viewers.

15. Choose a palette that matches data nature

Use qualitative palettes for categorical variables, sequential palettes for ordered numeric variables, and diverging palettes for data with a meaningful midpoint.

16. Design with accessibility in mind

Approximately one in twelve people are color‑blind; use varied saturation and brightness, and test charts in grayscale to ensure contrast.

17. Prioritize readability

Choose clear fonts, avoid italics, bold, all‑caps, and ensure high contrast with the background.

18. Use horizontal bar charts instead of rotated labels

When labels are long, keep them horizontal and use a bar chart to avoid rotating text, which hampers readability.

19. Choose a suitable chart library

For web and app projects, libraries such as ECharts or Highcharts provide built‑in interactions and follow the rules described.

20. Make charts interactive for user exploration

Allow users to adjust parameters, switch chart types, or modify timelines to explore data more effectively.

Conclusion

A well‑designed data visualization leaves a lasting mental model of facts, trends, or processes, while poor design adds unnecessary complexity.