ChatGPT for Data Analysis — Step-by-Step

Data Analysis Setup Prompt:

Dataset: Customer purchase history
Format: CSV with 50K rows
Columns: customer_id, purchase_date, amount, category, region

Goal: Identify high-value customer segments

Tools: Python, pandas, matplotlib, seaborn

I'll share the data structure and ask for analysis code.

Please write Python code that:
1. Loads the CSV file
2. Displays basic statistics (shape, dtypes, missing values)
3. Shows sample rows
4. Identifies outliers in the amount column
5. Checks date range

Dataset: customer_purchases.csv
Assume it's in the current directory.

import pandas as pd
import numpy as np

# Load data
df = pd.read_csv('customer_purchases.csv')

# Basic exploration
print(f"Shape: {df.shape}")
print(f"\nColumn types:\n{df.dtypes}")
print(f"\nMissing values:\n{df.isnull().sum()}")
print(f"\nFirst rows:\n{df.head()}")

# Statistics
print(f"\nBasic statistics:\n{df.describe()}")

# Outlier detection (using IQR)
Q1 = df['amount'].quantile(0.25)
Q3 = df['amount'].quantile(0.75)
IQR = Q3 - Q1
outliers = df[(df['amount'] < Q1 - 1.5*IQR) | (df['amount'] > Q3 + 1.5*IQR)]
print(f"\nOutliers: {len(outliers)} ({len(outliers)/len(df)*100:.2f}%)")

# Date range
print(f"\nDate range: {df['purchase_date'].min()} to {df['purchase_date'].max()}")

Data Cleaning Request:

Data issues:
- Some dates are in MM/DD/YYYY, others in YYYY-MM-DD
- Amount column has '$' and commas
- customer_id has leading zeros sometimes
- Some regions have extra spaces/capitalization
- Missing values in category column (about 2%)

Requested: Write code to clean all these issues

# Convert dates to consistent format
df['purchase_date'] = pd.to_datetime(df['purchase_date'], infer_datetime_format=True)

# Clean amount (remove $ and commas, convert to float)
df['amount'] = df['amount'].str.replace('$', '').str.replace(',', '')
df['amount'] = pd.to_numeric(df['amount'])

# Standardize customer_id (remove leading zeros)
df['customer_id'] = df['customer_id'].astype(int)

# Clean region (trim spaces, standardize case)
df['region'] = df['region'].str.strip().str.title()

# Handle missing categories (forward fill or default)
df['category'] = df['category'].fillna(method='ffill')
df['category'] = df['category'].fillna('Unknown')

# Remove duplicates if any
df = df.drop_duplicates()

Exploratory Data Analysis:

Dataset: Customer purchases (cleaned)
Columns: customer_id, purchase_date, amount, category, region

Create a comprehensive EDA that includes:
1. Distribution of purchases by category
2. Total revenue by region
3. Customer purchase frequency
4. Average order value by region
5. Trends over time
6. Correlation analysis if numeric columns exist

Provide code that generates summary statistics and visualizations.

Data Visualization Request:

Create 4 visualizations for customer purchase analysis:
1. Bar chart: Revenue by region
2. Line chart: Monthly revenue trend
3. Histogram: Distribution of purchase amounts
4. Scatter plot: Purchase frequency vs. average order value by customer

Use matplotlib and seaborn. Make them publication-ready
(good labels, colors, titles, font sizes).

import matplotlib.pyplot as plt
import seaborn as sns

# Set style
sns.set_style("whitegrid")
plt.rcParams['figure.figsize'] = (15, 10)

# 1. Revenue by region
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2)

revenue_by_region = df.groupby('region')['amount'].sum().sort_values()
revenue_by_region.plot(kind='barh', ax=ax1, color='steelblue')
ax1.set_title('Revenue by Region', fontsize=12, fontweight='bold')
ax1.set_xlabel('Total Revenue ($)')

# 2. Monthly revenue trend
df['month'] = df['purchase_date'].dt.to_period('M')
monthly_revenue = df.groupby('month')['amount'].sum()
ax2.plot(monthly_revenue.index.astype(str), monthly_revenue.values, marker='o')
ax2.set_title('Monthly Revenue Trend', fontsize=12, fontweight='bold')
ax2.tick_params(axis='x', rotation=45)

# 3. Distribution of purchase amounts
ax3.hist(df['amount'], bins=50, color='coral', edgecolor='black')
ax3.set_title('Distribution of Purchase Amounts', fontsize=12, fontweight='bold')
ax3.set_xlabel('Amount ($)')
ax3.set_ylabel('Frequency')

# 4. Scatter: Frequency vs Average Order Value
customer_stats = df.groupby('customer_id').agg({
    'customer_id': 'count',  # purchase count
    'amount': 'mean'  # average order value
}).rename(columns={'customer_id': 'purchase_count'})

ax4.scatter(customer_stats['purchase_count'], customer_stats['amount'], alpha=0.5)
ax4.set_title('Purchase Frequency vs Average Order Value', fontsize=12, fontweight='bold')
ax4.set_xlabel('Number of Purchases')
ax4.set_ylabel('Average Order Value ($)')

plt.tight_layout()
plt.savefig('analysis_dashboard.png', dpi=300, bbox_inches='tight')
plt.show()

Statistical Analysis:

Questions:
1. Is there a significant difference in average purchase amount between regions?
2. What's the correlation between purchase frequency and average order value?
3. Are certain categories more popular in specific regions?

Dataset: Customer purchases

Provide code to:
- Perform appropriate statistical tests (ANOVA, correlation, chi-square)
- Show p-values and effect sizes
- Interpret results

Customer Segmentation:

Using RFM analysis (Recency, Frequency, Monetary):
1. Calculate R: Days since last purchase
2. Calculate F: Number of purchases
3. Calculate M: Total amount spent

Create customer segments:
- High value: High R, F, M
- At-risk: High R, Low F, M
- Developing: Low R, Low F, Medium M
- VIP: Any R, High F and M

Provide code to segment customers and show segment size/statistics.

from datetime import datetime

# Reference date (today)
ref_date = pd.to_datetime('2024-12-31')

# Calculate RFM
rfm = df.groupby('customer_id').agg({
    'purchase_date': lambda x: (ref_date - x.max()).days,  # Recency
    'customer_id': 'count',  # Frequency
    'amount': 'sum'  # Monetary
})

rfm.columns = ['Recency', 'Frequency', 'Monetary']

# Quartile scores (1-4, 4 is best)
rfm['R_Score'] = pd.qcut(rfm['Recency'], 4, labels=[4, 3, 2, 1], duplicates='drop')
rfm['F_Score'] = pd.qcut(rfm['Frequency'].rank(method='first'), 4, labels=[1, 2, 3, 4])
rfm['M_Score'] = pd.qcut(rfm['Monetary'], 4, labels=[1, 2, 3, 4], duplicates='drop')

# Calculate RFM score
rfm['RFM_Score'] = rfm['R_Score'].astype(int) + rfm['F_Score'].astype(int) + rfm['M_Score'].astype(int)

# Segment
def segment(score):
    if score >= 10:
        return 'VIP'
    elif score >= 8:
        return 'High Value'
    elif score >= 5:
        return 'Medium Value'
    else:
        return 'Low Value'

rfm['Segment'] = rfm['RFM_Score'].apply(segment)

print(rfm['Segment'].value_counts())
print(f"\nSegment statistics:\n{rfm.groupby('Segment')[['Recency', 'Frequency', 'Monetary']].mean()}")

Sales Forecasting:

Using the monthly revenue data:
1. Plot monthly sales trend
2. Fit a linear regression to identify trend
3. Calculate seasonal patterns (monthly, quarterly)
4. Make a simple forecast for next 6 months

Provide code and interpretation.

Advanced Analysis Request:

Dataset: [describe data]
Business Question: [what you want to know]
Constraints: [any data limits]
Deliverables:
- Code for analysis
- Visualizations (specific types)
- Summary statistics
- Insights and recommendations

Assume I have pandas, numpy, matplotlib, seaborn, scipy installed.

Round 1: Initial exploration
"Analyze this customer dataset"
↓
ChatGPT provides: Basic exploration code and findings

Round 2: Deep dive
"The category X is growing fast. Analyze that trend deeper"
↓
ChatGPT provides: Focused analysis on category X

Round 3: Visualization
"Show category X trends by region with seasonal patterns"
↓
ChatGPT provides: Enhanced visualization code

Round 4: Business insights
"What actions should we take based on this analysis?"
↓
ChatGPT provides: Recommendations with supporting data