Data Visualization

Overview

• Turn analysis → clear stories fast.
• Tools over BigQuery feel “small data–speed” on billion+ rows (if model right).
• Some tools let you build charts w/o SQL (drag-drop), but the model still matters.

Perception

• Preattentive (<1s, 자동): color, position, length, size, orientation, shape, enclosure, intensity…

  → Use these to direct attention before people start “thinking”.

• Attentive (노력/working memory): we hold ~4 chunks max.

  → Keep visuals simple; one idea per chart. (한 차트 = 한 메시지)

Picking encodings

• Position/length > angle/area > color hue (정확도 순서).
• Use color sparingly (highlight 1–2 things).
• Sort bars; keep baselines at zero for bar charts. (막대는 0부터)

Choose effective charts

• Category 비교 → Bar (수평 권장).
  • Works with zeros, long labels.
• Part-to-whole (few slices) → Pie/Donut OK, but gets crowded fast.
• Time series → Line.
  • Don’t use line for static category split (gender by characters etc.).

80/20 reality check

• You spend 80% on wrangling, 20% on viz.

• Audience cares 99% about the final visuals.

  → Invest in layout, labeling, annotations, and a 1-line takeaway.

Dimensions vs Measures