Syndicated Data Storyteller

How to use it

1. 1Paste the prompt into ChatGPT, Claude or Gemini.
2. 2Paste your data table and fill the bracketed inputs (audience, business question).
3. 3Answer its clarifying questions about definitions and time periods.
4. 4Ask it to reframe the same story for a different audience.

See it in action

Story structure for Highland Creamery — 52-week cream cheese read:

Headline: Distribution grew but the new doors are underperforming — the share loss is a velocity problem concentrated in the two newest SKUs, not a brand problem.

Finding 1 — Core holds. The top four SKUs sit in your table's upper-right quadrant: share-stable, velocity above the category median. Say this first; it frames everything else as fixable.

Finding 2 — The drag is new-door velocity. The two recent launches gained distribution but sit bottom-left. The so-what: shelf presence without pull — a support and placement question, not a delist question yet.

Finding 3 — Price-gap timing. The share dip coincides with the competitor promo window visible in your table. Label the causal link an assumption to validate with promo-lift data.

Three charts worth building: the quad chart with quadrant labels renamed in plain language; a bridge from distribution gains to share change; new-SKU velocity trend vs the category median.

Actions supported: rebalance support toward the two launches, revisit their placement, pull the promo-overlap analysis before the next buyer meeting.

Want the two toughest leadership questions with suggested answers?

The prompt

Here's the start of the prompt. Download the free bundle for the full, ready-to-paste version — plus the installable Claude Skill and Custom-GPT instructions.

# Role
You are a senior category insights lead who has presented syndicated data to CEOs and skeptical buyers for fifteen years. You believe a table is not a story: you rank findings by money at stake, you separate what the data says from what it means, and you never stretch a number past what it can prove.

# Context I'll provide
- The data table(s), pasted as text: [PASTE YOUR TABLE — share, velocity, distribution, price, promo, any syndicated cut]
- Source and period: [SOURCE + TIME PERIOD e.g. Circana, 52 weeks ending June]
- The business question on the table: [BUSINESS QUESTION]
- Audience: [AUDIENCE e.g. leadership, retail buyer, sales meeting]
- What I already suspect (optional): [HYPOTHESES]

# Your task
1. If the table is unreadable, the period is unclear, or the business question is missing, ask up to 3 clarifying questions BEFORE writing anything.

Frequently asked questions

What is syndicated data in CPG?

Syndicated data is market measurement sold to many subscribers by providers like Nielsen or Circana: point-of-sale share, velocity, distribution, pricing, and promotion reads across retailers. It tells you what happened in the market. This skill turns those tables into a ranked narrative — findings, so-whats, charts, and actions — without inventing numbers beyond what you paste.

How do I paste a data table into the prompt?

Plain text works: copy cells straight from Excel — tab-separated columns are fine — or retype the key rows. Include headers, the time period, and units. If the model misreads a column, tell it what the column means and it will correct the reading before building the story.

Will this work in ChatGPT, Claude and Gemini alike?

Yes. The prompt is model-agnostic and runs in any capable chat model. Larger tables paste more cleanly into models with big context windows, but the storytelling structure is identical everywhere. Many teams save it as a Custom GPT or Claude Skill so monthly reporting follows one narrative standard.

What are the common mistakes when presenting syndicated data?

The classics: leading with the table instead of the finding, quoting share to two decimals as if precision were insight, mixing timeframes mid-story, and reading causation into a correlation the panel cannot prove. The skill counters each — it ranks findings, forces so-whats, keeps periods straight, and labels causal leaps as assumptions to validate.