Le Convertisseur de Données Brutes en Note Exécutive : Transformez des Chiffres Désordonnés en Analyse en 4 Étapes Prête à l'Action
Pourquoi ce prompt est important
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.
À quoi nous l'utilisons
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.Résultat
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.
La plupart des prompts « analyse ces données » produisent un résumé compétent de ce que disent les chiffres, ce qui n'est pas la même chose qu'une analyse sur laquelle un dirigeant peut agir. Un modèle à qui l'on demande d'« analyser ces données » décrira les tendances avec précision et s'arrêtera là, car rien dans une instruction vague ne l'oblige à faire le travail plus difficile d'interprétation.
Pourquoi quatre étapes explicites plutôt qu'une « analyse » ouverte
Séparer Tendances, Anomalies, Implications Commerciales et Actions Recommandées en étapes distinctes force le modèle à réellement accomplir chaque type de raisonnement au lieu de mélanger description et recommandation dans un seul paragraphe qui se lit bien mais ne tient pas la route.
Pourquoi la section Contraintes fait l'essentiel du vrai travail
L'exigence que « chaque affirmation doit référencer un chiffre spécifique » est ce qui empêche le modèle de se rabattre sur un langage vague comme « a montré une forte croissance ». L'instruction de signaler les problèmes de qualité des données avant d'en tirer des conclusions existe parce que les modèles traitent autrement chaque chiffre d'un jeu de données avec la même confiance.
Pourquoi les Implications Commerciales sont une étape séparée des Tendances
C'est l'étape que la plupart des analyses générées par IA sautent entièrement, et c'est celle qui compte vraiment pour le lecteur. « Le revenu de la Région X a augmenté de 41% » est une tendance. « La croissance de la Région X provient d'une base plus petite, donc ne lui allouez pas de ressources comme vous le feriez pour une hausse de 41% dans votre plus grande région » est une implication.
Comment en tirer le meilleur parti
Collez des données réelles, pas une description des données — le modèle a besoin des chiffres réels pour détecter des anomalies comme le pic du taux de retour dans l'exemple de sortie.