Visualizing Categories: Bar Charts vs Pie Charts

What is Categorical Data Visualization?

Bar Charts

Best for:

• Comparing quantities across categories
• Showing change over discrete time periods
• Ranking items

Key Properties:

• Bars have equal width with gaps between them
• Y-axis typically starts at zero
• Categories can be sorted for easier comparison

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# Sales data by product category
sales = pd.DataFrame({
    'Category': ['Electronics', 'Clothing', 'Food', 'Furniture', 'Sports'],
    'Revenue_M': [4.2, 2.8, 3.1, 1.9, 2.3],
    'Units_K': [85, 320, 450, 42, 180]
})

fig, axes = plt.subplots(1, 3, figsize=(15, 5))

# 1. Simple vertical bar chart
bars = axes[0].bar(sales['Category'], sales['Revenue_M'],
                    color=['steelblue', 'coral', 'mediumseagreen', 'orchid', 'orange'],
                    edgecolor='black', alpha=0.8)
axes[0].set_title('Revenue by Category')
axes[0].set_ylabel('Revenue ($M)')
axes[0].tick_params(axis='x', rotation=30)
# Add value labels on bars
for bar in bars:
    axes[0].text(bar.get_x() + bar.get_width()/2, bar.get_height() + 0.05,
                 f'${bar.get_height():.1f}M', ha='center', fontsize=9)

# 2. Horizontal bar chart (better for long labels)
sales_sorted = sales.sort_values('Revenue_M', ascending=True)
axes[1].barh(sales_sorted['Category'], sales_sorted['Revenue_M'],
             color='steelblue', edgecolor='black', alpha=0.8)
axes[1].set_title('Revenue by Category\n(Sorted — better for comparison)')
axes[1].set_xlabel('Revenue ($M)')

# 3. Grouped bar chart: multiple metrics
x = np.arange(len(sales))
w = 0.35
axes[2].bar(x - w/2, sales['Revenue_M'], w, label='Revenue ($M)', color='steelblue', alpha=0.8)
axes2b = axes[2].twinx()
axes2b.bar(x + w/2, sales['Units_K']/100, w, label='Units (100K)', color='coral', alpha=0.8)
axes[2].set_title('Grouped: Revenue vs Volume')
axes[2].set_xticks(x)
axes[2].set_xticklabels(sales['Category'], rotation=30)
axes[2].legend(loc='upper left')
axes2b.legend(loc='upper right')

plt.tight_layout()
plt.savefig('bar_charts.png', dpi=150)
plt.show()

Pie Charts

Best for:

• Part-to-whole relationships
• A small number of categories (≤ 5)
• When relative proportions are the main message

Worst for:

• Comparing values (bars are far better)
• Many categories (becomes unreadable)
• When precision matters

Mathematical Properties:

market_share = pd.DataFrame({
    'Company': ['Alpha', 'Beta', 'Gamma', 'Delta', 'Others'],
    'Share': [35, 28, 18, 12, 7]
})

fig, axes = plt.subplots(1, 2, figsize=(12, 5))

# Standard pie chart
colors = ['#2196F3', '#F44336', '#4CAF50', '#FF9800', '#9E9E9E']
wedges, texts, autotexts = axes[0].pie(
    market_share['Share'],
    labels=market_share['Company'],
    autopct='%1.1f%%',
    colors=colors,
    startangle=90,
    pctdistance=0.85
)
axes[0].set_title('Market Share (Pie Chart)')

# Donut chart (modern alternative)
wedges2, _, _ = axes[1].pie(
    market_share['Share'],
    labels=None,
    autopct='%1.1f%%',
    colors=colors,
    startangle=90,
    pctdistance=0.85,
    wedgeprops=dict(width=0.5)  # creates donut hole
)
axes[1].set_title('Market Share (Donut Chart)')
axes[1].legend(market_share['Company'], title='Company',
               loc='center left', bbox_to_anchor=(0.85, 0, 0.5, 1))

plt.tight_layout()
plt.savefig('pie_charts.png', dpi=150)
plt.show()

Bar Chart vs Pie Chart — Decision Guide

How many categories?
+-- ≤ 5 categories AND you want to show part-of-whole
|   +-- Pie chart (or donut) <- acceptable
|
+-- Otherwise -> Bar chart (almost always better)
    +-- Comparing values -> Vertical or horizontal bar
    +-- Many categories -> Horizontal bar (labels fit)
    +-- Change over time -> Line chart (not bar)
    +-- Multiple series -> Grouped or stacked bar

Common Mistakes

Bar & Pie Charts in Machine Learning

In ML, bar charts are everywhere:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split

# Generate dataset
X, y = make_classification(n_samples=1000, n_features=10, n_informative=5,
                           random_state=42)
feature_names = [f'Feature_{i}' for i in range(10)]

# Train model
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)

# Bar chart: feature importance
importance = model.feature_importances_
sorted_idx = np.argsort(importance)

plt.figure(figsize=(8, 5))
plt.barh(np.array(feature_names)[sorted_idx], importance[sorted_idx], color='steelblue')
plt.xlabel('Importance')
plt.title('Feature Importance (Random Forest)')
plt.tight_layout()
plt.show()

# Bar chart: class distribution
from collections import Counter
print("Class distribution:", Counter(y))

Key Takeaways

When in doubt, choose a bar chart. It's the safer, clearer option for almost every categorical comparison.