Data Storytelling Guide
Turn data into compelling stories that drive decisions.
Usage
- Start with the "So What?" — what action should the audience take based on this data?
- Structure as a narrative: context (what was happening) → tension (what changed/what's wrong) → resolution (what to do about it)
- Choose visualizations that make your point obvious without explanation
- Annotate charts to guide the viewer's eye to the key insight
- End with a clear recommendation, not just findings
Examples
- Before and after: Bad: "Q3 revenue was $2.4M." (So what?) Good: "Q3 revenue hit $2.4M, up 18% from Q2. But 80% of growth came from one enterprise deal. Excluding that deal, growth was 3% — below our 10% target. We need to diversify our pipeline." Same data, completely different narrative and action
- Chart annotation: Don't show a plain line chart and say "as you can see." Instead: draw a vertical line at the product launch date, add an arrow pointing to the inflection point, annotate: "Feature X launch: +23% activation." The chart tells the story without your narration
- The pyramid structure: Lead with the recommendation ("We should invest in mobile"). Then the supporting evidence ("Mobile users have 2x LTV, 40% of traffic, growing 15% monthly"). Then the details (methodology, caveats, data tables in appendix). Executives read the top. Analysts read the bottom. Everyone gets what they need
Guidelines
- Every data presentation should answer three questions: What happened? Why does it matter? What should we do?
- Use comparison, not absolute numbers: "Revenue is $2M" means nothing without context. "Revenue is $2M, 20% above target and the highest in company history" means everything
- One chart, one message. If a chart makes two points, split it into two charts
- Remove chartjunk: 3D effects, unnecessary gridlines, rainbow colors, legends that could be direct labels. Every pixel should convey information
- Know your audience: executives want implications and recommendations in 5 minutes. Analysts want methodology and data access. Tailor the same analysis for different audiences
- The "headline test": if someone only reads the title of each slide/section, do they get the story? Every title should be a finding, not a topic ("Mobile conversion increased 23%" not "Mobile metrics")