Data Visualization Speaking

Every day, executives, scientists, journalists, and students present data. Most do it BADLY — cluttered slides, unreadable charts, buried insights. But when data visualization is done well, it is one of the most powerful forms of communication. This lesson teaches how to pair DATA with NARRATIVE to inform and persuade.

Different data calls for different charts:\n\n- **Bar chart** — comparing categories (sales by region)\n- **Line chart** — change over time (stock price 2020-2026)\n- **Pie chart** — parts of a whole (use RARELY; humans are bad at comparing wedge sizes)\n- **Scatter plot** — relationships between two variables (height vs. weight)\n- **Heatmap** — patterns in 2D data (website click density)\n- **Histogram** — distribution of a single variable\n- **Sankey diagram** — flows (users through a funnel)\n\nThe single biggest mistake: using a pie chart for 10+ categories. No one can read it. Use a bar chart instead.

**1. ONE insight per chart.** If your chart has 4 messages, make 4 charts.\n\n**2. Headline title.** Title should state the INSIGHT, not the topic. Bad: "Sales by Quarter." Good: "Q4 Sales Surged 40% After Holiday Campaign."\n\n**3. Remove chart junk.** Gridlines, unnecessary labels, 3D effects, gradients — all distract. The great Edward Tufte calls this "data-ink ratio" — maximize ink showing data, minimize ink showing decoration.\n\n**4. Use color purposefully.** Highlight the key bar in a bold color; gray out the rest. Color should draw the eye to what matters.\n\n**5. Truncated axes lie.** Cutting the y-axis at 50 instead of 0 makes small differences look huge. Sometimes justified, but must be labeled honestly.

The single biggest leap in data storytelling: STOP reading numbers aloud and START telling stories around them.\n\nWeak: "Sales were 1.2M in Q1, 1.4M in Q2, 1.5M in Q3, 2.1M in Q4."\n\nStrong: "Take a look at Q4. You can see sales jumped 40% in just three months. This happened because we launched the holiday campaign on October 15 — that's this moment right here. (point at chart)"\n\nThe chart is the EVIDENCE. The story is the ARGUMENT. Do not force your audience to interpret the chart — TELL them what to see.

Structure for a data-heavy presentation:\n\n1. **State the punchline** — "Our new feature increased user retention by 15%."\n2. **Show the chart that proves it** — focus the audience's eye on the key change.\n3. **Explain the why** — what caused the effect?\n4. **Address caveats** — "This excludes the power-user cohort because..."\n5. **State the implication** — "So we are rolling out to all users next month."\n\nThis pattern is called "BLUF" (Bottom Line Up Front) and is used by journalists, consultants, military briefers, and top executives.

- **Tableau / Power BI** — industry standard for dashboards\n- **Python (matplotlib, seaborn, plotly)** — for analysts and scientists\n- **R (ggplot2)** — academic research gold standard\n- **D3.js / Observable** — custom web visualizations (NYT / FT use this)\n- **Datawrapper** — journalist-friendly, free tier\n- **Flourish** — no-code interactive charts\n- **Figma / PowerPoint** — for polishing and custom annotation\n\nThe TOOL matters less than the principles. A beautiful bar chart in PowerPoint beats a bad heatmap in D3.

- **Hans Rosling** — Gapminder's "200 Countries Over 200 Years" TED talk. Masterclass in animated storytelling.\n- **Edward Tufte** — classic books "The Visual Display of Quantitative Information"\n- **Nathan Yau** — Flowing Data blog\n- **The New York Times Graphics** — gold standard for journalism viz (COVID maps, election forecasts)\n- **Financial Times Visual Journalism** — uses charts as arguments\n- **Alberto Cairo** — "The Truthful Art," academic + practitioner\n\nStudying these builds your eye. When you notice their tricks, you can apply them.

Data viz is persuasive. With that comes responsibility. Avoid:\n\n- Truncated y-axes without noting it\n- Cherry-picked time windows\n- Area charts with squared sizes (people read lengths, not areas)\n- 3D charts that distort comparison\n- "Colorblind-unfriendly" color schemes\n- Implying causation from correlation\n\nThe best data storytellers always ask: is this chart HONEST? Would a skeptical colleague agree?