Good Prompts vs. Bad Prompts in Data AnalysisA Practical Guide for Analysts, Business Leaders & AI Practitioners

Introduction

Artificial intelligence has fundamentally changed how data analysis is performed.

Tools like Claude, ChatGPT, and Gemini can now:

• Clean datasets

• Build dashboards

• Write SQL queries

• Generate forecasts

• Explain KPIs

However, the quality of output depends entirely on the prompt.

A vague prompt produces unreliable results.A structured prompt produces accurate, actionable insights.

Core Principle

A great prompt includes:

1. Context — What is the data

2. Task — What you want

3. Constraints — Format, tools, audience

4. Output — Desired result

5. 
   

SECTION 1 · DATA CLEANING (Examples 1–5)

Example 1

Bad Prompt:Clean my data.

Improved Prompt:Detailed dataset description, issues, tools, and expected output.

Example 2

Bad Prompt:Remove outliers.

Improved Prompt:Defines method (IQR), outputs, and explanation requirements.

Example 3

Bad Prompt:Fix the dates.

Improved Prompt:Specifies formats, target structure, and validation rules.

Example 4

Bad Prompt:Deduplicate my data.

Improved Prompt:Defines exact duplicates and fuzzy matching logic.

Example 5

Bad Prompt:Analyze my data.

Improved Prompt:Requests full data quality audit with structured output.

SECTION 2 · DATA VISUALIZATION (Examples 6–10)

Bad Prompt:Make a chart.

Improved Prompt:Provides:

• Data

• Tool (Chart.js)

• Visual requirements

• Design style

• 
  

SECTION 6 · ADVANCED ANALYSIS (Examples 22–25)

Example 22 — Dashboard Design

Defines:

• Data model

• Page structure

• Metrics

• Visuals

• DAX measures

• 
  

Example 23 — Anomaly Detection

Uses:

• Z-score

• Seasonal decomposition

• Rolling IQR

Includes:

• Severity rating

• Visualization

• 
  

Example 24 — Cohort Analysis

Defines:

• Dataset structure

• Weekly cohorts

• Retention matrix

• Engagement metrics

• 
  

Example 25 — Executive Summary

Specifies:

• Business metrics

• Writing framework

• Word count

• Decision-focused output

Conclusion & Key Takeaways

Across all examples:

The difference between weak and powerful AI results is prompt quality.

Best Practices

1. Provide real data or schema

2. Specify tools (Python, SQL, Power BI)

3. Define business context

4. Specify output format

5. Add constraints

6. Request explanations

7. Include validation steps

8. Break complex tasks into steps

   
   

Final Insight

AI amplifies your team’s capability — but only if you know how to communicate with it effectively.