Day 04 Part 2 — Data Storytelling & Presentation¶
The best analysis in the world is worthless if nobody acts on it. Data storytelling is the skill that turns findings into decisions — structuring a narrative, choosing the right visuals, and communicating to people who don't think in data. It's the single most career-defining skill for an analyst.
Learning Objectives¶
- Structure an analytical narrative (situation → insight → action)
- Tailor communication to the audience
- Annotate charts to direct attention to the insight
- Handle questions and pushback
- Avoid the most common presentation mistakes
Why Storytelling Matters¶
The hard truth
"The analyst who communicates a good-enough analysis clearly will always beat the analyst who communicates a brilliant analysis poorly."
Stakeholders don't reward correctness — they reward clarity. Your job isn't to show your work; it's to change a decision.
The Narrative Structure¶
Every data story follows the same arc:
1. SITUATION — set the context. What's the business question?
2. COMPLICATION — what's the problem or surprise the data reveals?
3. INSIGHT — the key finding, stated plainly
4. EVIDENCE — the chart/number that proves it
5. ACTION — what should we do about it?
Example:
"We're growing revenue 8% a year (situation). But margin is falling (complication). The cause is discounting above 20%, which loses money on every sale (insight). Here's the chart showing profit turning negative at high discounts (evidence). I recommend a 20% discount cap, which could recover £130k in profit (action)."
That's a complete data story in five sentences.
Know Your Audience¶
| Audience | Cares about | How to present |
|---|---|---|
| Executives | The decision, the bottom line | Lead with the answer, minimal detail, one chart |
| Managers | What to do with their team | Specific actions, segment-level data |
| Analysts/Technical | The method, the rigour | Methodology, assumptions, caveats |
| Cross-functional | How it affects them | Translate to their world, avoid your jargon |
The same analysis needs different presentations
A churn finding for the CEO ("we're losing £8M/year to churn, here's the fix") is completely different from the same finding for the data team ("here's the model, features, and validation"). Don't give executives the technical version.
Chart Annotation — Direct the Eye¶
A chart shows data. An annotation shows the point.
# Don't just show the trend — annotate the insight
ax.annotate(
"Discounts above 20% lose money",
xy=(2, -28000), xytext=(0.5, -60000),
arrowprops=dict(arrowstyle="->", color="red"),
fontsize=11, fontweight="bold", color="red"
)
Techniques: - Title states the finding, not the chart name ("Discounting Above 20% Destroys Profit", not "Profit by Discount Band") - Highlight the key data point in colour; mute everything else in grey - Add a reference line (target, average) for context - Annotate the "so what" directly on the chart
(Full coverage in Week-01/Day-05-Part-1-Data-Visualization-Basics/05-visualization-best-practices.)
Slide Design Principles¶
- One message per slide. If a slide has three points, it has zero.
- The title carries the message. A reader who only reads titles should get the whole story.
- One chart per slide. Don't make people choose where to look.
- Remove everything that isn't the point — logos, decoration, redundant labels.
- End with the recommendation, not a "thank you" slide.
Handling Questions and Pushback¶
- "How confident are you?" — Have your caveats ready. Honesty builds trust: "This is based on 3 months of data; I'd want to validate with a test before fully committing."
- "That doesn't match my numbers." — Don't get defensive. Reconcile: "Let's check — are we using the same date range and definitions?" (See the Sales Dashboard interview answer.)
- "What should we do?" — Always have a recommendation. "I don't know" is rarely the right answer; "Based on this, I'd recommend X, while we test Y" is.
- Silence/skepticism — Anchor back to the business impact in money or customers.
Common Presentation Mistakes¶
The data dump
Showing every chart you made. The audience doesn't care about your journey — only the destination. Cut ruthlessly.
Burying the lead
Building up to the insight at the end. State it first; executives stop listening once they have what they need.
No recommendation
Presenting findings without a "so what." Analysis without a recommended action is homework, not consulting.
Jargon
"The p-value was 0.03 so we rejected the null." Say: "We're confident this difference is real, not luck."
Practice Exercises¶
Warm-up 1. Take a finding from any course project and write it as a 5-sentence data story (situation → complication → insight → evidence → action). 2. Rewrite a chart title from descriptive ("Revenue by Month") to insight-driven. 3. Translate this for an executive: "The cohort retention curve plateaus at 38% after month 3, with a hazard rate concentrated in the first 30 days."
Main 4. Build a 5-slide executive presentation from one of the projects: situation, two findings, recommendation, next steps. 5. Annotate a chart from a project to highlight its key insight with title, colour, and an arrow. 6. Prepare answers to three likely pushback questions for that presentation.
Interview Questions¶
[Beginner] Why is communication considered the most important analyst skill?
[Mid-level] How would you present the same churn finding differently to a CEO vs a data science team?
[Mid-level] A stakeholder says your numbers don't match theirs. How do you handle it in the moment?
[Senior] Walk me through how you'd structure a 10-minute presentation of a complex analysis to a skeptical executive audience.