data-storytelling

Overview of data-storytelling skill

What the data-storytelling skill does

The data-storytelling skill helps turn raw analysis into decision-ready communication. Instead of stopping at charts, it pushes you to structure findings as a narrative with context, tension, insight, and action. This is most useful when your real job is not just “analyze data,” but “explain what matters and what should happen next.”

Best fit for report writing and stakeholder communication

This data-storytelling skill is a strong fit for analysts, operators, consultants, founders, and report authors who need to explain metrics to executives or non-technical readers. It is especially relevant for data-storytelling for Report Writing, quarterly reviews, board updates, investor decks, and recommendation memos.

What makes this different from a generic prompt

A normal prompt may summarize numbers. The data-storytelling skill gives a reusable communication frame: setup, conflict, resolution; hook, context, rising action, climax, resolution, call to action; and the balance of data, narrative, and visuals. That structure reduces the common failure mode where analysis is accurate but forgettable.

What users should care about before installing

The key adoption question is simple: do you need persuasive explanation or just analysis output? If your audience must understand significance, tradeoffs, and next steps, this skill adds value. If you only need SQL help, chart generation, or raw metric summaries, it is probably too high-level on its own.

What is included in the repository

This skill is lightweight. The repository evidence shows a single SKILL.md with the main guidance and no extra scripts, rules, or references. That means setup is easy, but output quality depends heavily on the quality of the prompt and the data context you provide.

How to Use data-storytelling skill

data-storytelling install context

Install the skill from the parent repository, then invoke it from your AI workflow:

npx skills add https://github.com/wshobson/agents --skill data-storytelling

Because this skill lives at plugins/business-analytics/skills/data-storytelling, you are effectively pulling a narrative framework for analytics communication rather than a runnable analytics toolchain.

Read this file first

Start with:

• SKILL.md

There are no supporting README.md, rules/, resources/, or helper scripts surfaced for this skill, so nearly all practical value is in understanding and applying the framework inside SKILL.md.

What input the skill needs to work well

The data-storytelling skill performs best when you provide more than a dataset dump. Give it:

• the audience: executive, client, manager, board, investor, customer
• the format: report, memo, slide outline, brief, spoken update
• the decision to support: invest, cut, prioritize, diagnose, explain
• the core metrics and time window
• the baseline or comparison point
• the most important finding
• known limitations or uncertainty
• the desired action

Without those inputs, the model can still produce a story, but it may sound polished while missing business relevance.

Turn a rough goal into a strong prompt

Weak goal:

• “Write a data story about churn.”

Stronger data-storytelling usage prompt:

• “Use the data-storytelling skill to turn this churn analysis into a report section for a VP of Customer Success. Audience is non-technical. Goal is to justify retention investment for Q3 planning. Use setup-conflict-resolution. Start with a strong hook, explain the baseline, identify the most important driver, quantify business impact, note confidence limits, and end with 3 recommended actions.”

This improves output because it defines the audience, decision, structure, and expected ending.

Recommended prompt template

Use a template like this for consistent results:

• Objective: what this report needs to achieve
• Audience: who will read it
• Data: key numbers, trends, comparisons
• Context: what changed and why it matters
• Constraints: tone, length, format, certainty
• Output request: narrative structure, visuals to suggest, recommendations

Example:

• “Apply the data-storytelling skill. Write a 500-word executive summary for a quarterly business review. Data: revenue +8% QoQ, gross margin -3 pts, churn concentrated in SMB accounts, CAC rising 12%. Context: leadership deciding whether to shift budget from acquisition to retention. Include hook, context, rising action, key insight, recommendation, and next steps.”

Suggested workflow for report writing

A practical workflow for data-storytelling for Report Writing:

1. Extract the few metrics that matter.
2. Identify the tension: decline, gap, risk, opportunity, surprise.
3. Choose the decision-maker audience.
4. Ask the model to draft a narrative arc.
5. Review whether the “climax” is truly the most decision-relevant insight.
6. Add visual recommendations only after the storyline is stable.
7. Trim anything that does not support the main decision.

This order matters. Many weak reports start with too many charts and discover too late that there is no clear story.

How to choose the right framework

The source skill emphasizes a few durable structures. In practice:

• Use Setup → Conflict → Resolution when you need a crisp memo or report section.
• Use the longer narrative arc when you need a presentation or executive walkthrough.
• Use the three pillars—data, narrative, visuals—when the draft feels imbalanced.

A good install decision test: if your team repeatedly has analyses that are “interesting but not actionable,” this skill is likely worth adopting.

What good inputs look like

Better inputs are comparative and decision-linked, for example:

• “Conversion dropped from 4.2% to 3.1% after pricing changes”
• “Enterprise renewals offset SMB churn, masking segment risk”
• “Support backlog rose 28% while NPS fell 6 points”
• “The business choice is whether to hire support staff or reduce onboarding friction”

These are stronger than isolated numbers because they create tension and explain why someone should care.

Common usage mistakes

Most weak data-storytelling usage comes from one of these mistakes:

• asking for “a compelling story” without naming the audience
• providing metrics without a baseline
• skipping the recommendation
• overloading the narrative with every observed pattern
• forcing causal claims from descriptive data
• treating visuals as decoration instead of evidence support

The skill works best when you narrow the message to one central insight and one clear action path.

How this skill fits with normal analysis work

The data-storytelling skill does not replace analysis, data cleaning, or chart production. It sits after those steps. A strong workflow is: analyze first, then use the skill to package findings into a narrative that survives executive skim-reading.

What output to ask for

Useful output requests include:

• executive summary
• board-ready memo
• quarterly review narrative
• investor update section
• slide-by-slide outline
• insight-to-action brief
• annotated chart captions

If you ask only for “a story,” you often get style without decision utility. Ask for a business document type instead.

data-storytelling skill FAQ

Is the data-storytelling skill good for beginners?

Yes, if you already have the data or findings. The framework is simple and approachable. Beginners may still struggle with choosing the single most important insight, so it helps to explicitly ask the model to rank findings by business impact before drafting.

When should I use this instead of a normal summarization prompt?

Use the data-storytelling skill when the audience needs persuasion, context, and a recommended action. Use a normal summary prompt when you only need a factual recap of results.

Is this skill only for presentations?

No. It is equally useful for reports, memos, executive emails, quarterly reviews, and investor-facing writeups. The core value is narrative structure, not slides specifically.

Does data-storytelling install include charts or automation?

No evidence suggests built-in scripts, chart tooling, or automation in this skill. data-storytelling install gives you a communication framework, not a visualization engine or reporting pipeline.

Can I use it for technical audiences?

Yes, but it is most valuable for mixed or non-technical audiences. For deeply technical readers, you may want a more direct structure with lighter narrative framing and more methodological detail.

When is this skill a poor fit?

Skip this skill when:

• you still have not validated the analysis
• the audience only wants raw tables or technical appendix detail
• the decision is trivial and does not need persuasion
• you need domain-specific statistical rigor more than communication structure

How is it different from a slide-writing skill?

A slide-writing skill focuses on format and presentation flow. The data-storytelling guide here is about shaping evidence into meaning first. You can apply it before writing slides, reports, or spoken remarks.

How to Improve data-storytelling skill

Start with the decision, not the dataset

The fastest way to improve data-storytelling output is to define the decision the story must support. “Summarize this dashboard” is weak. “Help leadership decide whether churn warrants retention investment” is much stronger.

Provide tension explicitly

Stories need a conflict. If your prompt does not include one, the model may invent drama or produce bland prose. Name the tension directly:

• growth with declining margin
• higher revenue with worsening retention
• segment gains hiding segment losses
• improving top-line metrics with rising operational risk

Rank insights before drafting the story

Before asking for the final narrative, ask the model to do this first:

• identify the top 3 findings
• rank them by business significance
• select one as the central message
• explain which decision it should influence

This prevents the common problem where the first draft tries to tell five stories at once.

Add baselines and comparisons

Comparisons make narratives credible. Improve your data-storytelling guide inputs with:

• previous period vs current period
• target vs actual
• segment vs segment
• before vs after intervention
• internal trend vs market benchmark

A story without comparison often reads like description, not insight.

Control the level of certainty

One major failure mode is overstating what the data proves. Tell the model whether findings are descriptive, directional, or causal. Ask it to separate:

• what the data shows
• what is likely driving it
• what needs further validation

This increases trust, especially in executive or investor settings.

Ask for visuals only after the narrative is stable

The source skill values visuals, but charts should support the story rather than lead it. A useful iteration sequence is:

1. get the hook and key message right
2. validate the conflict and evidence
3. tighten recommendations
4. then ask which chart best clarifies each point

Improve report-writing outputs with section constraints

For data-storytelling for Report Writing, specify section behavior:

• opening sentence must state the business stakes
• context paragraph must define the baseline
• evidence section must use only 3 supporting points
• recommendation section must include owner, timing, and expected impact

These constraints materially improve usefulness because they force actionability.

Fix outputs that sound polished but empty

If the first draft feels generic, revise with one or more of these instructions:

• “Use the exact numbers provided.”
• “Name the affected segment explicitly.”
• “State the tradeoff behind the recommendation.”
• “Cut any claim not supported by the data.”
• “Replace abstract language with operational implications.”
• “End with a concrete next step.”

Iterate from narrative quality, not just wording

Do not only edit tone. Evaluate whether the draft has:

• a clear hook
• enough context for the audience
• one memorable key insight
• a believable recommendation
• a next step someone can act on

If one of those is missing, the issue is structure, not phrasing.

Build a reusable house style around data-storytelling

If your team writes recurring reports, create a standard prompt wrapper around the data-storytelling skill with fixed fields for audience, decision, metrics, baseline, risk, confidence, and recommendation. This reduces variability and makes the skill more reliable across recurring business reviews.