Performing a Complete EDA Workflow

A structured EDA workflow combines data inspection, statistical analysis, group comparisons, and visualization to answer a specific question.

Consider a retail dataset containing product, region, quantity, price, and customer rating.

Step 1: Prepare the Dataset

import pandas as pd

import matplotlib.pyplot as plt

df = pd.DataFrame({

"Product": ["Laptop", "Phone", "Tablet", "Laptop", "Phone", "Tablet"],

"Region": ["North", "South", "East", "South", "North", "East"],

"Quantity": [10, 25, 15, 12, 30, 18],

"Price": [55000, 30000, 25000, 55000, 30000, 25000],

"Rating": [4.5, 4.2, 4.0, 4.7, 4.4, 3.9]

})

df["Revenue"] = df["Quantity"] * df["Price"]

print(df.head())

The Revenue column is a derived variable calculated by multiplying quantity by price. Creating useful derived metrics often helps answer business questions more effectively.

Step 2: Summarize Overall Performance

print("Total revenue:", df["Revenue"].sum())

print("Average rating:", df["Rating"].mean())

print("Total units sold:", df["Quantity"].sum())

These metrics provide an overview of the dataset's overall performance.

Step 3: Compare Products and Regions

product_summary = df.groupby("Product").agg(

Total_Revenue=("Revenue", "sum"),

Total_Quantity=("Quantity", "sum"),

Average_Rating=("Rating", "mean")

)

print(product_summary)

This summary compares products across several measures. A product with the highest revenue may not have the highest quantity sold because prices differ.

Step 4: Visualize the Results

product_summary["Total_Revenue"].plot(

kind="bar",

title="Total Revenue by Product",

xlabel="Product",

ylabel="Revenue",

rot=0

)

plt.tight_layout()

plt.show()

The chart makes it easier to compare revenue across products than reading the raw records alone.

Step 5: Interpret the Findings

When interpreting the results, consider questions such as:

• Which product contributes the most revenue?
• Does a product sell many units but generate less revenue?
• Are ratings similar across product categories?
• Are any values missing, inconsistent, or unusually high?
• Would additional data help explain the observed patterns?

EDA should produce evidence-based observations rather than unsupported conclusions. Documenting the findings also helps guide the next steps, such as selecting visualization techniques, performing statistical tests, or preparing data for machine learning.