Grouping, Summaries, and Visualisation

What is a grouped analysis?

Grouping follows a split–apply–combine pattern:

1. Split records into groups (e.g., by region)
2. Apply a calculation to each group (e.g., sum of revenue)
3. Combine the results into a summary table

Use it to compare sales by region, average delivery time by courier, or sign-ups by marketing channel.

Example

summary = (df.groupby('region', as_index=False)
.agg(total_revenue=('revenue', 'sum'),
order_count=('order_id', 'nunique'),
avg_order_value=('revenue', 'mean'))
.sort_values('total_revenue', ascending=False))
print(summary)

How it works

Common aggregation functions: 'sum', 'mean', 'median', 'count', 'nunique', 'min', 'max', 'std'

Other useful summaries:

Create a chart

import matplotlib.pyplot as plt
plt.figure(figsize=(8, 4))
plt.bar(summary['region'], summary['total_revenue'])
plt.title('Total revenue by region')
plt.xlabel('Region')
plt.ylabel('Revenue (₹)')
plt.tight_layout()
plt.show()

Always label axes, show units, and use a chart type that makes the comparison easy.

Keeping notebooks reproducible:

• Put all imports in the first cell.
• Run Restart & Run All before sharing, to confirm the notebook works top to bottom.
• Use relative file paths and record library versions (pd.__version__).
• Avoid editing data by hand outside the code.

Common mistakes

• Wrong unit of analysis: a line-item table and an order-level table give different counts. Use nunique on order IDs.
• count vs nunique: count counts non-null rows; nunique counts distinct values.
• Unlabelled charts: no title, units, or axis labels.
• Grouping on dirty categories: clean labels before grouping, or "South" and "south " become separate groups.

Try it: Group a sales dataset by region, calculate total revenue, and create a bar chart of the results.