A good chart tells a story instantly. A bad chart buries the story under poor design choices. The difference isn't about the data — it's about how you present it. Here are the principles that separate effective data visualisation from decorative clutter.
Know Your Audience
The same data needs different visualisations for different audiences. An executive dashboard should show high-level KPIs with trend lines and clear targets. A data scientist exploring a dataset needs detailed scatter plots, histograms, and the ability to drill down. Design for who will read it, not for yourself.
Start with the Message
Before choosing a chart type, ask: what's the one thing the reader should take away? If the answer is "sales are growing faster in Q4 than any previous quarter," design the chart to make that obvious. A line chart with Q4 highlighted, a callout annotation, and clear axis labels does this. A table of monthly sales numbers does not.
Use Colour Purposefully
Colour should encode meaning, not decorate. Each colour should represent a category, a value range, or a state. Random colour assignment makes charts harder to read, not easier.
- Use one accent colour to highlight the important data point, and grey for the rest
- For sequential data (low to high), use a sequential colour scale (light to dark)
- For diverging data (negative to positive), use a diverging scale (red to green)
- Avoid red and green together for the 8% of men with colour vision deficiency — use blue and orange instead
Label Everything Clearly
Every chart needs:
- A descriptive title that states the main finding, not just the topic ("Q4 Sales Grew 32% Year-over-Year" beats "Monthly Sales Data")
- Axis labels with units ("Revenue (₹ Lakhs)" not just "Revenue")
- Direct labels on data points where possible, instead of relying on a legend
- A source line if the data comes from external research
Remove Chart Junk
Edward Tufte coined the term "chart junk" for visual elements that don't add information: gridlines that are too dark, 3D effects, gradient fills, drop shadows, decorative backgrounds. Every visual element competes for attention with your data. Remove anything that doesn't serve the message.
Specific things to remove:
- Chart borders and backgrounds
- Heavy gridlines (use light, thin gridlines or none at all)
- 3D effects on 2D data
- Gradient fills on bars or lines
- Excessive tick marks on axes
- Legends when direct labels work better
Choose the Right Aspect Ratio
Chart width-to-height ratio affects perception. Line charts with too much vertical space exaggerate small changes. Bar charts that are too wide make comparisons harder. A 16:9 or 4:3 ratio works for most charts. For long time series, use a wider ratio. For few data points, use a squarer ratio.
Make Charts Accessible
- Provide a text alternative (alt text or a data table) for screen readers
- Don't rely on colour alone — add patterns, labels, or textures to distinguish categories
- Ensure sufficient contrast between data elements and background
- Make charts keyboard-navigable for interactive visualisations
- Test with a colour blindness simulator before publishing
Annotations Guide the Reader
Don't make the reader figure out the story. Add annotations directly on the chart:
- Callout boxes for key data points
- Reference lines for averages, targets, or thresholds
- Shaded regions for important time periods (recessions, product launches)
- Text labels on outliers explaining what they are
Responsive Design
Charts need to work on mobile. On small screens:
- Simplify — remove non-essential elements
- Use shorter labels or rotate them
- Consider stacking vertical charts instead of side-by-side
- Make sure tooltips work with touch, not just hover
The chart tool on this site supports responsive rendering out of the box. Charts automatically resize to fit their container. Paste your data, pick a chart type, and customise. All processing happens in your browser.