TutorialsSelf-Practice Templates & Pitfall Guide
8 timeline-ordered prompt templates (data exploration → sales analysis → expense analysis → funnel → survey → chart styling → action recommendations → QA checklist), plus 3 general red lines + 4 data-analysis-specific red lines.
8 Self-Practice Templates
Ordered by usage timeline: data exploration → sales analysis → expense analysis → funnel → survey → chart styling → action recommendations → QA checklist. Each template can be copied directly — just replace the [placeholders].
Template 1: Data Exploration (Data Profiling)
Don't rush to analyze. First, scan [filename] and report:
How many columns, what each column means, date range, total rows,
any null values, duplicate records, or anomalous numbers,
whether formats are consistent.
Report findings only — don't start analyzing.[Placeholder: filename] | Constraint: report only, no analysis
Template 2: Sales Analysis
This is [business context, e.g.: company Q1 2026 sales detail, fields include
date, salesperson, product, quantity, amount, province, customer type].
I want to see [question, e.g.: regional and product sales performance].
Help me analyze:
1. Overall performance: total sales, total order count (by unique order ID),
average order value; calculate achievement rate if targets exist
2. Time trends: monthly comparison, identify peaks and valleys
3. Products: sales Top 10 and share
4. Regions: comparison and ranking
Definition: [e.g., tax-inclusive, orders by unique ID].
Output: HTML chart report + key conclusions[Placeholders: business context / question / definition] | Constraint: order count by unique ID; targets cannot be fabricated
Template 3: Expense Analysis
Analyze [expense record file]:
1. Department cost rankings
2. Expense type proportions (travel/entertainment/office/training)
3. Monthly cost trends
4. Flag anomalous expenses (threshold > 5000)
5. Calculate approval rate
Output: HTML bar chart report; list anomalous items: date/employee/department/
type/amount/approval status/record location
Note: Items above threshold are flagged "for review" only — don't classify as violation.[Placeholders: filename / threshold] | Constraint: anomaly ≠ violation; above threshold = "for review" only
Template 4: User Behavior Funnel
Analyze [user behavior data file]:
1. Calculate conversion rates at each stage (home/product page/cart/checkout/payment)
2. Find the biggest drop-off stage
3. View activity by hour
4. Compare new vs. returning user conversion rates
Output: HTML funnel chart + auxiliary charts; conclusions under 300 words
Note: Confirm definition first — by unique users or by events; if you see
"cart > checkout" illogical funnels, check data issues first,
don't treat as normal conversion.[Placeholders: filename / funnel stages] | Constraint: verify definition first; non-standard funnels → check data issues
Template 5: Satisfaction Survey
Analyze [survey file]:
1. Per-question averages and score distributions
2. Find the 3 lowest-scoring dimensions (priority improvement items)
3. Sentiment analysis on open-ended questions; positive/neutral/negative ratios
4. Top 3 most frequent keywords in negative feedback
5. Compare satisfaction by age/city/purchase frequency
Note: With only a few dozen responses, don't present conclusions as representative
of all customers; flag uncertain sentiment classifications for manual review.[Placeholders: filename / question range] | Constraint: small samples → no overgeneralization; sentiment classifications → spot-check raw responses
Template 6: Chart Styling
Redo charts with this style: colors [corporate blue primary, green/orange secondary],
title and body font sizes [specified], overall [clean business style],
hover tooltips show data, responsive without blurring,
each chart has title/axes/legend/unit/data source.[Placeholders: colors/style] | Constraint: don't just say "generate charts" — be specific
Template 7: Action Recommendations
Based on the analysis results above:
1. Distill 3 most important findings, written as "finding → recommendation"
2. Create an A4 one-page management dashboard for management
3. Propose 5 immediately actionable improvement measures, ranked by priority
For each: which data supports it, who is responsible, deadline,
how to measure effectiveness.
No vague phrases like "strengthen management" or "continuous optimization" —
if unclear, keep asking me.[Placeholders: number of findings / dashboard format] | Constraint: chase down vague phrases, convert to actionable items
Template 8: QA Checklist (9 Items)
Go through the analysis results against these 9 items, respond "pass/fail + reason" for each:
1. Is the data source and version correct?
2. Are field types and missing values handled?
3. Are metric definitions, denominators, and filter conditions clear?
4. Do chart numbers match the source table and body text?
5. Is correlation incorrectly stated as causation?
6. Are anomalies directly classified as violations?
7. Are sample conclusions overgeneralized?
8. Do recommendations have data support and assigned owners?
9. Is sensitive data desensitized and access controlled?[Placeholder: paste analysis results] | Constraint: go through all 9 items; failures must be explained
7 Quality Red Lines
General Red Lines (3)
1. AI fabricates with confidence: Numbers, dates, and names must be verified manually — when AI is wrong, it sounds just as assertive.
2. Never feed sensitive information: Desensitize before input, or don't input at all — once sent, you can't take it back.
3. Output is not the final draft: Your name is on it, you're responsible — review everything before sending.
Data Analysis Specific Red Lines (4)
4. Anomaly ≠ violation: Above threshold means "needs further review," not a conclusion.
5. Correlation ≠ causation: When AI says "causes," ask for evidence.
6. Don't overgeneralize from samples: A few dozen surveys don't represent all customers.
7. Metric definitions must be clear; targets cannot be AI-fabricated: Definitions come from you, not from AI guessing.
9-Step Workflow Hands-On
Step by step — from data preparation to report review, run the full data analysis pipeline using a single sales dataset to produce a complete analysis report on the spot.
Analyze a Dataset with AI
Take a dataset, run the 9-step workflow, and produce a one-page management dashboard + action recommendations.