Grouping and Aggregations (groupby, agg)

The idea

groupby answers questions like:

• Average revenue per city?
• Total orders per customer?
• Highest score per class?

Think: split → apply → combine.

From the book

McKinney credits Hadley Wickham for the term split-apply-combine: Pandas splits your rows into groups by key, applies a function (like sum or mean) to each group independently, then combines the per-group results back into one Series or DataFrame. Calling .groupby("city") alone does nothing yet — it just remembers the grouping until you call an aggregation like .mean().

Example dataset

Sales sample

import pandas as pd
 
sales = pd.DataFrame({
    "city": ["pune", "pune", "delhi", "delhi", "delhi"],
    "category": ["A", "B", "A", "A", "B"],
    "amount": [100, 200, 150, 120, 300],
})
 
print(sales)

Basic aggregations

Sum per city

Sum per city

print(sales.groupby("city")["amount"].sum())

Mean per city

Mean per city

print(sales.groupby("city")["amount"].mean())

Multiple aggregations with agg

Multiple metrics

summary = (
    sales.groupby("city")
    .agg(
        total_amount=("amount", "sum"),
        avg_amount=("amount", "mean"),
        orders=("amount", "count"),
    )
)
 
print(summary)

Group by multiple keys

Group by city and category

summary = (
    sales.groupby(["city", "category"])
    .agg(total_amount=("amount", "sum"), orders=("amount", "count"))
    .reset_index()
)
 
print(summary)

Common gotchas

• After groupby, the grouped columns become the index. Use .reset_index() if you want them as normal columns.
• Aggregation functions ignore missing numeric values by default (NaN) in many cases.

Mental model

When you write:

python

sales.groupby("city")["amount"].sum()

You are saying:

• Split the rows by city
• Select the amount column
• Sum within each group

This pattern comes up constantly in analytics.

Visualize it

diagram Split, apply, combine mermaid

  flowchart LR
  A["Original rows"] --> B["Split by key (city)"]
  B --> C1["Group: pune"]
  B --> C2["Group: delhi"]
  C1 --> D1["Apply sum()"]
  C2 --> D2["Apply sum()"]
  D1 --> E["Combine into result"]
  D2 --> E

groupby splits rows into groups, applies a function to each group independently, then combines the results into one table.

sketch Split → apply → combine p5.js Pause Restart

groupby splits rows by key, applies an aggregation to each group separately, then combines the answers into one table.

🧪 Try It Yourself

Exercise 1 – Group and Sum

pch.exerciseTag pch.exerciseTitle

Exercise 2 – Multiple Aggregations With agg

pch.exerciseTag pch.exerciseTitle

Exercise 3 – Reset the Index After Grouping

pch.exerciseTag pch.exerciseTitle

Next

groupby and agg cover the essentials — continue to Advanced GroupBy (transform, filter, named agg) for broadcasting group stats back, filtering whole groups, and bucket analysis.