Prompt Template for Thorough Exploratory Data Analysis:

I want you to do a comprehensive, exploratory data analysis on [this data/topic].

Process requirements:

Create a notes.md file to document findings as you go
Maintain a running TODO list of questions to investigate - and ADD to it when you discover interesting threads worth pulling
Use Python scripts heavily to explore the data - don't just answer questions, look for patterns, anomalies, and surprises
When you find something unexpected, pursue it. The goal is discovery, not just answering a checklist.
Do NOT produce any final report or summary until you've exhausted your exploration

Mindset:

Be a detective, not a report-writer
"That's weird" is a signal to dig deeper, not move on
Breadth first, then depth on the interesting stuff
Minimum 7-10 phases of analysis before synthesizing

Only when you're truly done exploring, summarize what you found most insightful. Then we'll discuss how to structure a final deliverable.

The key phrases that shift your behavior:

"Add to your TODO list when you discover threads" — prevents you from treating it as a fixed checklist
"Do NOT produce any final report until..." — explicitly blocks your tendency to prematurely deliver
"The goal is discovery, not just answering" — reframes success criteria
"'That's weird' is a signal to dig deeper" — legitimizes exploratory tangents
"Minimum 7-10 phases" — sets an expectation of depth

Reusable Prompt for Data Analysis Report Generation

DOCUMENT STRUCTURE

1. Title Page

Project/client name prominently displayed
Descriptive subtitle (what was analyzed)
Date
Summary metrics table (5-7 key numbers that frame the scope)

2. Top Insights Section (60% of document focus)

This is the heart of the report. For each major insight:

Format each insight as:

Numbered heading with a punchy, conclusive title (not a question—state the finding)
1-2 paragraph explanation in plain language (no jargon, no statistical notation)
A chart OR simple table that visually proves the point
Brief methodology nod (1-2 sentences, non-technical): "This was measured by comparing X across Y" or "Based on analysis of N projects over Z time period"
So-what statement: why this matters or what action it implies

Chart/Table Guidelines for Top Insights:

Prefer charts over tables when showing comparisons or trends
Tables should be 2-4 columns max, no more than 6-8 rows
Every visual should have an obvious takeaway within 3 seconds
Label notable data points directly on charts
Use color meaningfully (red=bad, green=good, blue=neutral)

Methodology Nods (Top Section):

Keep it to one sentence
Use plain English: "We compared..." not "A Mann-Whitney U test was performed..."
Focus on what was measured, not how
Example: "Based on grouping the 34 projects by their naming patterns and comparing average productivity"

3. Detailed Analysis Section (30% of document focus)

This section supports the top insights with technical depth. Include:

For each analysis area:

Section heading matching a theme (e.g., "Tier Analysis," "Predictive Modeling")
Brief context paragraph
Detailed tables with full statistics
Methodology boxes (technical) with:
- Exact formulas or calculations used
- Statistical tests performed with test statistics and p-values
- Sample sizes and data filters applied
- Assumptions and limitations

Methodology Box Format: Use a visually distinct shaded box labeled "📊 Methodology: [Topic]"

Bullet points with technical details
Include actual numbers: "n=34, p=0.006, R²=0.262"
Note any caveats: "3 projects excluded due to missing data"

4. Recommendations Section (10% of document focus)

Numbered list of actionable recommendations
Each item: bold action statement + supporting rationale in regular text
Quick reference table for practical application (estimation lookup, risk checklist, etc.)

STYLISTIC PREFERENCES

Tone:

Confident and conclusive (not hedging)
Insight-first, evidence-second
Written for a smart non-technical reader

Formatting:

Professional blue color scheme
Consistent heading hierarchy
Generous white space
Alternating row shading in tables
Page breaks before major sections

Charts:

Clean, minimal style (no 3D, no excessive gridlines)
Direct labels on data points for key callouts
Legends only when necessary
Title states the conclusion, not just the topic
- Good: "Rules Projects Are 15x Slower Than Average"
- Bad: "Velocity by Project Family"

Tables:

Header row with dark background, white text
Right-align numbers, left-align text
Include units in headers, not cells
Bold key figures that support the insight

Language:

Avoid: "It should be noted that...", "It is interesting that...", "As can be seen..."
Prefer: Direct statements. "X is Y." "The data shows Z."
Quantify everything: "significantly faster" → "3.2x faster"

WHAT TO AVOID

Dense paragraphs without visual breaks
Tables with more than 5-6 columns
Methodology in the top section that reads like a statistics textbook
Insights that bury the conclusion at the end of a paragraph
Charts that require explanation to understand
Repeating the same data in both chart and table form
Executive summary that's just a list of section headings

EXAMPLE INSIGHT BLOCK (for reference)

3. Small Projects Take MORE Time, Not Less

Counter-intuitively, the smallest projects took longer than the largest ones. Projects in the bottom 25% by size averaged 71 days, while projects in the top 25% averaged just 55 days—despite being 29x larger.

[CHART: Side-by-side bar comparison of Bottom 25% vs Top 25% showing size and days]

This pattern reflects a fixed overhead of approximately 50 days per project (environment setup, testing, deployment, documentation) that dominates small project timelines. Based on comparing the smallest and largest project quartiles.

Implication: Consider consolidating small APIs into larger migration units where architecturally feasible.

Apply this structure to generate the report for: [DESCRIBE YOUR ANALYSIS AND KEY FINDINGS]

Quick Reference: The Formula

Section	Audience	Visuals	Methodology Depth	Length
Title Page	Everyone	Summary table	None	1 page
Top Insights	Executives	Charts + simple tables	1 sentence plain English	60%
Detailed Analysis	Technical readers	Detailed tables	Full methodology boxes	30%
Recommendations	Decision makers	Lookup tables	None	10%

Optional Add-On Prompts

If you want more charts:

For each insight, prefer a chart over a table. Use bar charts for comparisons, scatter plots for relationships, and line charts for trends. Every chart title should state the conclusion.

If you want a specific number of insights:

Identify the top [N] most surprising or actionable findings. Prioritize findings that challenge conventional assumptions or have clear business implications.

If you have a specific audience:

The primary audience is [ROLE]. Tailor the language complexity and focus areas accordingly. [ROLE] cares most about [SPECIFIC CONCERNS].

If you want anomaly callouts:

Include a section on "Anomalies Requiring Investigation" that lists data points that don't fit the patterns, with specific questions to explore.

DroopyTersen/data-exploration.prompt.md

Select an option

No results found