matplotlib & seaborn Basics for Data Analysts

Analysis nobody can see is analysis nobody acts on. As a data analyst in India, your numbers become decisions only when you can show them clearly - and the two libraries that do the heavy lifting in Python are matplotlib and seaborn. This guide gives you the core charts, the rule for choosing between them, and the clean-styling habits interviewers notice.

How the two libraries relate

• matplotlib is the foundational plotting engine. It gives you full control over every element but is verbose.
• seaborn is built on top of matplotlib. It produces attractive statistical charts with far less code and sensible defaults.

The practical workflow: use seaborn for quick, good-looking statistical plots, and drop down to matplotlib when you need fine control over labels, titles, or layout. They cooperate - seaborn returns matplotlib axes you can keep tweaking.

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

sns.set_theme(style="whitegrid")   # instantly cleaner defaults

The figure and axes model

Interviewers like the explicit object-oriented style over scattered plt. calls. Create a figure and an axes, then draw on the axes:

fig, ax = plt.subplots(figsize=(8, 4))
ax.plot([1, 2, 3], [10, 20, 15])
ax.set_title("Sample")
plt.show()

fig is the whole canvas; ax is the plotting area. This pattern scales cleanly to multiple subplots and is the professional default.

Core charts and when to use each

Line chart - trends over time

Use a line chart when the x-axis is continuous, especially dates. "Zomato revenue by month" is the classic case:

revenue = pd.DataFrame({
    "month": ["Jan", "Feb", "Mar", "Apr", "May"],
    "amount": [120000, 135000, 128000, 156000, 171000],
})

fig, ax = plt.subplots(figsize=(8, 4))
ax.plot(revenue["month"], revenue["amount"], marker="o")
ax.set_title("Monthly Revenue (Rupees)")
ax.set_xlabel("Month")
ax.set_ylabel("Revenue")
plt.show()

Bar chart - comparing categories

Use bars to compare discrete categories - revenue per city, orders per restaurant. seaborn makes this one line:

city_sales = pd.DataFrame({
    "city": ["Mumbai", "Delhi", "Pune", "Jaipur"],
    "sales": [820000, 760000, 540000, 390000],
})

sns.barplot(data=city_sales, x="city", y="sales")
plt.title("Sales by City (Rupees)")
plt.show()

Histogram - distribution of one numeric variable

Use a histogram to see how values spread - are most cab fares small with a few large ones?

fares = pd.Series([120, 150, 90, 320, 200, 110, 480, 175, 130, 260])
sns.histplot(fares, bins=8)
plt.title("Distribution of Cab Fares (Rupees)")
plt.show()

Scatter plot - relationship between two numerics

Use scatter to check whether two variables move together - does order amount relate to delivery distance?

sns.scatterplot(data=orders, x="distance_km", y="amount")
plt.title("Order Amount vs Delivery Distance")
plt.show()

Box plot - distribution across groups, with outliers

A box plot compares spreads side by side and exposes outliers - ideal for "fare distribution per city":

sns.boxplot(data=rides, x="city", y="fare")
plt.title("Cab Fare Distribution by City")
plt.show()

A quick chart-choice cheat sheet

• Trend over time -> line
• Compare categories -> bar
• Distribution of one number -> histogram
• Relationship between two numbers -> scatter
• Distribution across groups / spot outliers -> box plot

Choosing the wrong chart is the most common mistake - a line chart over unordered cities, or a bar chart where a histogram belongs. Match the chart to the question.

Clean styling habits that signal competence

Recruiters can tell a polished chart from a default one. Build these habits:

• Always label. Title, x-label, y-label - a chart without them is unreadable. Use set_title, set_xlabel, set_ylabel.
• Size sensibly. Set figsize so bars are not cramped and text is legible.
• Rotate crowded ticks. plt.xticks(rotation=45) when city or product names overlap.
• Use sns.set_theme() once at the top for consistent, professional aesthetics.
• Call plt.tight_layout() before showing or saving to stop labels being clipped.
• Save crisp images with plt.savefig("chart.png", dpi=150, bbox_inches="tight") for reports.

seaborn extras worth knowing

• hue splits any plot by a category with automatic colours: sns.barplot(data=df, x="city", y="sales", hue="category").
• sns.countplot is a bar chart of category frequencies - "number of orders per city" without pre-aggregating.
• sns.heatmap visualises a correlation matrix or a pivot table - excellent for "sales by city x month".

Common pitfalls

• Forgetting plt.show() in scripts - the figure renders but never appears.
• Overplotting. Thousands of points in a scatter become a blob; sample or add transparency with alpha=0.3.
• Mixing plt. and ax. carelessly. Pick the axes-based style and stay consistent.
• No labels. The fastest way to look like a beginner.

Visualization is where your analysis earns its keep. Learn these five charts cold, label everything, and let seaborn handle the polish.

Related: apply, map & lambda in pandas (and When to Avoid Them) · Practise Python

Don't just read. Prove your skill on DevWithData.