designmotionhq

visual

Charts That Lie

Same data, opposite stories: how you draw a chart decides which truth people see.

Watch on Instagram

Key insights

  • Start every bar chart's y-axis at zero. Truncating the baseline turns a +4% change into a fake +400% explosion — the single most common way charts mislead.
  • Match the chart to the question. Bars compare values, lines show change over time, and pie charts fall apart past ~5 slices. The question picks the form, not your taste.
  • Aspect ratio rewrites the trend. The same rising series looks flat when squished and like a spike when stretched — balance it so the average slope sits near 45° and reads honestly.
  • Maximize the data-ink ratio. Strip gridlines, drop shadows, 3D skew, and boxed legends, then label the line directly. Every remaining pixel should carry data.
  • Color is encoding, not decoration. Use one hero color to spotlight the series that matters, and choose categorical, sequential, or diverging scales to fit the data type.
  • Title the chart with the takeaway, not the metric. "Revenue flat since March" tells the story; "Quarterly revenue" makes the reader hunt for it.

Do / Don't

  • Do: Start every bar chart's y-axis at zero, no exceptions.
  • Do: Pick the chart type from the question you're answering.
  • Do: Spotlight the one series that matters with a single accent color.
  • Don't: Truncate or crop an axis to exaggerate small differences.
  • Don't: Add gridlines, shadows, or 3D effects that encode no data.
  • Don't: Reach for a pie chart when you have more than five slices.

Want Claude to build this correctly?

UX Engine gives Claude Code the senior-designer reasoning behind every pattern.

The Design System Blueprint

An 11-page visual guide to consistent, scalable UI: colors, typography, spacing, components, motion.

Get the Blueprint — free

The PDF, plus emails about new patterns and my tools. Unsubscribe anytime.

Related patterns