The Raw Data to Executive Brief Converter: Turn Messy Numbers Into a 4-Stage Action-Ready Analysis
Why this prompt matters
Raw data doesn't drive decisions on its own — a specific, numbers-backed narrative built from that data does. Analysts who hand over spreadsheets or generic dashboards without translating numbers into business implications routinely watch their findings get ignored, because executives rarely have time to do that translation themselves. An analyst who can't produce a clear, defensible recommendation from the data gets treated as a report generator rather than a strategic partner, and their insights get deprioritized behind whoever tells the more compelling story — even when that story is built on weaker data.
What we use it for
You're a data analyst at a mid-size e-commerce company. Q3 sales data just landed in a spreadsheet with dozens of columns across 5 regions, and your VP of Sales wants the 3 things that matter by end of day — not a dashboard, not a slide deck, just the insights.
Prompt
Role: You are a data analyst and business intelligence consultant with deep experience translating raw datasets into board-level insights for [YOUR INDUSTRY].
Context: I have the following raw data: [PASTE DATA HERE — CSV, table, or list of numbers]. This data covers [TIME PERIOD / SCOPE, e.g., "Q3 2026 sales by region and product line"]. The audience for this analysis is [AUDIENCE, e.g., "the VP of Sales, who has 10 minutes and no patience for raw numbers"].
Task: Analyze this data in four stages: (1) identify the 3-5 most significant trends, (2) flag any anomalies or outliers with a plausible explanation for each, (3) translate the findings into business implications, (4) recommend 2-3 specific actions ranked by expected impact.
Constraints: Do not simply describe the data ("sales went up") — explain why it matters and what changed. Every claim must reference a specific number from the data, not a vague trend. Flag any data quality issues (missing values, suspicious outliers, inconsistent units) before drawing conclusions from them. Keep the entire output under [MAX WORD COUNT, e.g. 400] words — this is for an executive, not a report.
Output Format: Structure as four labeled sections — Trends, Anomalies, Business Implications, Recommended Actions. Use bullet points, not paragraphs. Each bullet must be a single sentence a reader could act on without re-reading the raw data.Result
Filled in for: Q3 2026 e-commerce sales data by region (North America, EMEA, APAC, LATAM, MEA), covering revenue, units sold, and return rate, prepared for a VP of Sales.
Trends
• North America revenue hit $4.2M this quarter, up 18% QoQ — the largest absolute contributor to overall growth, driven mainly by the new bundle pricing launched in July.
• APAC grew 41% QoQ to $1.9M, the fastest percentage growth of any region, off a smaller base than North America or EMEA.
• EMEA revenue was flat at $2.8M (+3% QoQ), suggesting the region has plateaued after two consecutive quarters of double-digit growth.
• Overall units sold grew 22% company-wide, faster than the 15% revenue growth — average order value is declining slightly as growth skews toward lower-priced SKUs.
Anomalies
• LATAM revenue dropped 12% QoQ to $640K while units sold were flat — this is a price/mix problem, not a demand problem, and warrants checking whether a currency shift or a promotional discount is compressing margin.
• MEA grew 65% QoQ, but on a base of only $310K — statistically eye-catching, practically small; do not present this figure without the base number attached, or it will be misread as a major trend.
• LATAM's return rate is 22%, nearly triple the company average of 8% — this is worth flagging as a data quality or fulfillment issue before it's read as a demand signal.
Business Implications
• The July bundle pricing change is doing real work in North America and should be considered for EMEA, where growth has stalled at a similar starting point to where North America was before the change.
• The declining average order value means the 22% unit growth is not translating proportionally into revenue — worth investigating whether promotional activity is training customers toward lower-value purchases.
• LATAM's combination of falling revenue and elevated returns suggests a fulfillment or product-fit issue specific to that region, not a broader demand weakness the other numbers would support.
Recommended Actions
• Highest impact: pilot the North America bundle pricing model in EMEA next quarter, given the similar market conditions before North America's change.
• Medium impact: audit LATAM's fulfillment process and return reasons before the next planning cycle — the 22% return rate is large enough to be actively eroding the region's contribution margin.
• Lower impact but low-cost: hold MEA's growth rate out of headline reporting until the region crosses a materiality threshold (e.g., $1M quarterly), to avoid resourcing decisions based on a statistically noisy small-base number.
Most "analyze this data" prompts produce a competent summary of what the numbers say, which is not the same thing as an analysis an executive can act on. A model told to "analyze this data" will describe trends accurately and stop there, because nothing in a vague instruction forces it to do the harder work of interpretation.
This prompt is built around a specific failure mode: data summaries that are accurate but useless, because they never cross the line from description into implication.
Why four explicit stages instead of one open-ended "analyze"
Separating Trends, Anomalies, Business Implications, and Recommended Actions into distinct stages forces the model to actually complete each type of reasoning rather than blending description and recommendation into a single paragraph that reads well but doesn't hold up. Anomaly detection in particular gets skipped entirely by vague prompts — a model asked to "summarize the data" will often present a small-base percentage spike as equivalent in importance to a large-base trend, exactly the kind of misread that erodes trust in data-driven recommendations.
Why the Constraints section does most of the real work
The requirement that "every claim must reference a specific number" is what prevents the model from defaulting to vague language like "showed strong growth." The instruction to flag data quality issues before drawing conclusions from them exists because models otherwise treat every number in a dataset as equally trustworthy — an outlier caused by a data entry error gets analyzed with the same confidence as a genuine trend unless explicitly told to check for that first.
The word count constraint is doing quieter but equally important work: it forces prioritization. An unconstrained analysis will report everything it noticed: A tightly bounded one has to decide what actually matters enough to make the cut, which is the same judgment call a human analyst has to make before walking into an executive's office.
Why Business Implications is a separate stage from Trends
This is the step most AI-generated analyses skip entirely, and it's the one that actually matters to the reader. "Revenue in Region X grew 41%" is a trend. "Region X's growth came from a smaller base, so don't resource it the same way you'd resource a 41% increase in your largest region" is an implication — and it's the sentence that actually changes what the VP does next.
How to get the most out of this
Paste real data, not a description of the data — the model needs the actual numbers to catch anomalies like the return-rate spike in the example output. If the first pass surfaces an anomaly you can't explain, ask a follow-up question narrowing in on just that anomaly rather than re-running the full four-stage analysis; it's a faster way to get a defensible explanation before you present the findings.