Data Storytelling SKILL.md

This is the exact file agents read, including the credit header and licence notice. Save it as data-storytelling/SKILL.md.

---
name: data-storytelling
description: "Turn data into stories people act on: context, one insight per visual, evidence, explanation, implication and action, finding and checking insights, choosing the right chart, decluttering and highlighting, headline titles and annotations, clear uncertainty, common data traps and recommendations. Use whenever someone needs to present data, explain metrics or results, make charts that make a point, write commentary for a dashboard, report experiment results or turn analysis into a clear recommendation."
license: CC-BY-4.0
metadata:
  author: Afeez Adebayo
  author_url: https://afeezadebayo.com
  author_linkedin: https://www.linkedin.com/in/afeez-adebayo/
  author_x: https://x.com/za_icon
  source: https://afeezadebayo.com/skills-md/productivity/communication-presentations/data-storytelling/
  category: productivity
  subcategory: communication-presentations
  version: "1.0.0"
  created: "2026-10-14"
  updated: "2026-10-14"
---

# Data Storytelling

Numbers rarely persuade on their own. A dashboard full of accurate charts can leave a room unsure what happened, why it matters or what to do. Data storytelling turns analysis into a clear message: one insight, the evidence that proves it, the context that gives it meaning and the action it supports. Cole Nussbaumer Knaflic's Storytelling with Data popularised a practical approach: understand the context, choose an effective visual, remove clutter, focus attention and tell a story. [Reference, Storytelling with Data] Brent Dykes describes data stories as the combination of data, narrative and visuals. [Reference, Effective Data Storytelling]

This skill turns data into stories people act on: finding the insight, framing it for the audience, choosing the chart that proves it, decluttering and highlighting, writing headline titles and annotations, adding context and comparisons, handling uncertainty openly and ending with a recommendation. It works for slides, reports, dashboards, memos and spoken updates.

## Operating principles
1. **Insight before chart.** Decide what the data says before deciding how to show it.
2. **One message per visual.** Each chart proves one point stated in its title.
3. **Context makes numbers meaningful.** Compare with targets, past periods, benchmarks or peers.
4. **Focus attention.** Highlight the data that matters; mute everything else.
5. **Show uncertainty plainly.** Sample sizes, ranges and limitations build trust.
6. **End with action.** What should the audience decide or do because of this?

## The data story structure
```
1. Context: what we looked at and why it matters to this audience
2. Insight: the one thing the data shows (headline)
3. Evidence: the chart or numbers that prove it
4. Explanation: why it happened (drivers, causes, segments)
5. Implication: what it means (money, customers, risk, time)
6. Action: what we recommend, with owner and date
```

## Finding the insight
| Question | Example |
|---|---|
| What changed? | Sign-ups fell 18% in August |
| Compared with what? | Versus July, versus target, versus last year |
| Where is it concentrated? | Mostly mobile users from paid social |
| What explains it? | A pricing page change on 4 August |
| So what? | About 900 fewer trials a month, roughly £40,000 pipeline |
| Now what? | Restore the old layout for mobile and test the new one |

An insight states a change or difference, its size, where it happens and why it matters. "Sign-ups were 4,120" is a fact; "Mobile sign-ups fell 18% after the pricing change, costing about 900 trials a month" is an insight.

## Choosing the visual
| Message | Chart | Avoid |
|---|---|---|
| Change over time | Line chart | Many lines without highlighting |
| Compare categories | Horizontal bar, sorted | Unsorted bars, 3D |
| Part of a whole | Stacked bar or a table with percentages | Pies with many slices |
| Distribution | Histogram, box plot | Averages alone |
| Relationship | Scatter plot with a highlighted group | Dual axes that mislead |
| Ranking | Sorted bar or slope chart | Alphabetical order |
| Before and after | Slope chart or paired bars | Separate charts on different scales |
| Single key figure | Big number with comparison | A chart for one number |
| Exact lookup | Simple table | Charts when values must be read precisely |

## When to use, and when not to
Use it for performance reviews, board and leadership updates, research findings, experiment results, dashboards that need commentary, investor updates and any presentation that relies on numbers.

Don't use it to do the analysis itself (statistics, modelling) or to present weak data as certain; if the data cannot support a conclusion, say so.

## Inputs
- The question the data answers and the audience.
- The data, analysis and definitions (metrics, periods, segments).
- Targets, benchmarks or past periods for comparison.
- The format: slide, report, dashboard, memo or spoken update.
- Known limitations: sample sizes, missing data, changes in tracking.

## Workflow

### Step 1. Define the audience and decision
Who will see it and what will they decide? A CFO needs cost and risk; a product team needs user behaviour; a board needs trends and outlook.

### Step 2. Explore and find the insight
Look for changes, differences, concentrations and outliers. Ask "compared with what?" and "so what?" until the insight is clear.

### Step 3. Check the insight
Is the change real (not noise or a tracking change)? Is the sample big enough? Are there other explanations? Confirm definitions.

### Step 4. Write the headline
One sentence with the change, size and meaning: "Small business churn doubled in Q3, driven by billing problems in the first 60 days."

### Step 5. Choose the visual
Use the table above; pick the simplest chart that proves the headline.

### Step 6. Declutter
Remove gridlines, borders, unnecessary legends, decimals and colours. Label data directly.

### Step 7. Focus attention
Use one accent colour for the data that proves the headline; grey everything else. Add one or two short annotations at the key point.

### Step 8. Add context
Targets, previous periods, benchmarks or a reference line; explain units and definitions in a small note.

### Step 9. Explain and recommend
Drivers in one or two sentences, the implication in money, customers or time and a specific recommendation.

### Step 10. State limitations
Sample size, confidence and what would change the conclusion.

## Before and after
**Before:** a chart titled "Monthly sign-ups by channel" with seven coloured lines and a legend.

**After:** title "Mobile paid-social sign-ups fell 18% after the 4 August pricing change"; the mobile paid-social line in the accent colour, others grey; a vertical marker at 4 August labelled "new pricing page"; a note "about 900 fewer trials a month"; source and period below.

## Annotation rules
- Annotate the point that proves the headline, not every point.
- Use plain words: "Pricing page changed" rather than "Event A".
- Keep annotations to a few words; detail goes in speaker notes or text.

## Handling uncertainty
| Situation | How to show it |
|---|---|
| Small sample | State n; use cautious wording ("early signs") |
| Estimates and forecasts | Ranges or shaded bands; state assumptions |
| A/B test results | Effect size with confidence interval; avoid "proved" |
| Tracking changes | Mark the date; do not compare across the break without adjustment |
| Correlation | Say "associated with", not "caused", unless tested |

## Common data traps
- Truncated axes that exaggerate bar differences.
- Cherry-picked periods that hide the trend.
- Averages that hide skewed distributions.
- Percentages without base numbers ("up 200%" from 2 to 6).
- Comparing segments of very different sizes without normalising.
- Simpson's paradox: a trend that reverses when data is split by group.

## Edge cases and judgement calls
| Situation | What to do |
|---|---|
| No clear insight | Say so; show what was checked and what more data would help |
| Bad news in the data | Lead with it plainly, with causes and the plan |
| Executive audience | Headline numbers, one chart per point, recommendation first |
| Technical audience | More method and distribution detail, still headline-led |
| Dashboard that nobody reads | Add a short written summary of what changed and why |
| Conflicting data sources | Show both, explain differences, state which is used and why |

## Handoffs to related skills
- Presentation structure: `presentation-storyline`
- Slide layouts: `slide-design`
- Delivering the numbers: `presenting-delivery`
- Written reports: the Workplace Writing `reports` skill
- Executive summaries: the Workplace Writing `executive-summaries` skill
- GTM and growth reporting: the GTM Engineering `gtm-reporting` and Growth Marketing `growth-reporting` skills

## Worked example
A product team's monthly review showed eight dashboards and no conclusion. Exploring the data, the agent found that mobile sign-ups from paid social fell 18% after a pricing page change on 4 August, while desktop was stable. It checked tracking (no change), sample size (over 20,000 visits) and seasonality (no similar dip last year). The story became one slide: headline, a line chart with mobile paid social highlighted and a marker on 4 August, an annotation of about 900 fewer trials a month, an implication of roughly £40,000 pipeline at risk and a recommendation to restore the old mobile layout and test the new one properly. The team approved the change in the meeting, and sign-ups recovered within three weeks.

## Common mistakes
- Showing data without saying what it means.
- Chart titles that name the metric instead of the insight.
- Too many series, colours and labels.
- No comparison, so numbers have no meaning.
- Claiming causation from correlation.

## Output format
```
1. Audience and decision
2. Insight headline
3. Evidence: chart type, highlight and annotations
4. Context and comparisons
5. Explanation and implication
6. Recommendation with owner and date
7. Limitations and confidence
```

## Self-check before you answer
- Is there one insight per visual, stated in the title?
- Is the data compared with something meaningful?
- Is attention focused on the data that proves the point?
- Are limitations stated, and does the story end with an action?

---
Created by Afeez Adebayo, Marketing, SEO & Growth Expert
Website: https://afeezadebayo.com
LinkedIn: https://www.linkedin.com/in/afeez-adebayo/
X: https://x.com/za_icon
Source: https://afeezadebayo.com/skills-md/productivity/communication-presentations/data-storytelling/

Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this skill, including for commercial use, as long as you credit Afeez Adebayo and link to the source URL above. Keep this notice in any copy or adapted version.