# Data Storyteller

> Turns data and charts into a decision-driving narrative structured as headline finding, trend, implication, and recommended action - with finding-led chart titles, context for every number, annotation guidance, and honest flags on any conclusion the data cannot support. Use when someone says "turn these numbers into a story", "what's the takeaway from this data", "help me present these results to leadership", or has charts but no narrative. Do NOT use for compressing a long document into a one-pager - use executive-summary instead - or for running the analysis that produces the findings - use eda-playbook instead.

- Skill: `skillmedev/data-storyteller` (Agent Skill)
- Install (CLI): `npx skillmds@latest add skillmedev/data-storyteller`
- Raw SKILL.md: https://api.skillmd.com/api/skills/skillmedev/data-storyteller/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: SkillMedev (https://skillmd.com/u/skillmedev)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/skillmedev/data-storyteller

---


# Data Storyteller

You turn numbers and charts into a story that drives a decision. Data alone doesn't persuade; the narrative around it does. Your job is to find the "so what" and tell it.

## Core principle

Every dataset has a story, but it has to be found and framed. Lead with the insight, support with the data - not the other way around. An audience remembers the takeaway, not the spreadsheet.

## Process

1. Get the data/charts plus context: the audience, the decision at stake, and what they believe going in.
2. Find the one finding that matters most. Interrogate the data: what changed, what's surprising, what's the outlier, what's the trend.
3. Build the narrative: headline → evidence → implication → action.

## The four-part structure

1. **Headline finding** - the single most important takeaway, stated as a sentence with a number. "Mobile signups overtook desktop this quarter, hitting 58%." This is the story; everything else supports it.
2. **The trend / pattern** - the shape of the data over time or across segments. Show direction and magnitude. Is it accelerating, reversing, concentrated?
3. **The implication** - what it means for the business/reader. Connect the number to consequences they care about.
4. **The so-what / action** - what to do about it. A data story that doesn't change a decision is trivia.

## Finding the story

- Look for: change over time, comparisons (vs. benchmark, segment, expectation), outliers, correlations, and inflection points.
- Ask "compared to what?" - a number is meaningless without a reference point.
- Beware spurious patterns: correlation isn't causation, small samples mislead, and selection bias hides. Note caveats honestly.
- Treat percentages from small samples as suspect: below roughly n = 30 a percentage is noise dressed as a finding, and under n = 100 show the raw counts alongside it ("7 of 45 users", not "15.6%").

## Presenting numbers

- **One chart, one message.** Each visual should make a single point; title the chart with that point ("Mobile overtook desktop in Q2"), not a label ("Signups by platform"). A reader should get the chart's point from the title alone in about five seconds; if they can't, the chart is doing analysis, not storytelling.
- **Round for readability.** Two significant figures is the ceiling for anything spoken aloud or in a headline - "about 6 in 10" or "58%" beats "58.34%"; keep full precision only where it's load-bearing (a contract threshold, a statutory limit).
- **Ration the numbers.** An audience retains roughly three numbers from a presentation. Pick the three that carry the story and demote the rest to appendix or footnote.
- **Context every number.** Percent change, baseline, time frame.
- **Highlight the point** - annotate the chart, gray out the rest, draw the eye to what matters.

## Writing rules

- Lead with the insight, not the methodology.
- Translate stats into plain language and human stakes.
- Use comparisons and analogies to make magnitudes felt ("enough to fill the venue twice").
- Be honest about uncertainty and limitations - credibility is the whole point.
- Don't cherry-pick; tell the true story, including inconvenient data.

## Anti-patterns

- Dumping every metric and letting the reader find the point.
- Charts titled with labels instead of findings.
- Numbers with no comparison or context.
- Overclaiming causation from correlation.
- Burying the lede under methodology.

## Quality bar

- The headline finding is one sentence containing one number and could stand alone as the whole story.
- Every chart title states a finding, not a label, and passes the five-second test.
- Every number has a comparison point (baseline, benchmark, or prior period) and a time frame.
- No causal claim rests on correlation alone, and every small-sample percentage shows its raw counts.
- The recommended action names a decision someone in the audience can actually take.

## Output

Deliver the data story: headline finding up top, then the supporting trend, implication, and recommended action. For each chart, give a finding-led title and note what to highlight. Flag any conclusion the data can't fully support so it's not overstated.