Matplotlib & Seaborn

Charts are how analysts *show* findings. Matplotlib is the foundational Python plotting library; Seaborn sits on top and adds statistics + nicer defaults. Together they cover 95% of static charts you'll ever need.

What Matplotlib Actually Is

Matplotlib draws a Figure (the canvas) which contains one or more Axes (each plot). Inside an Axes are Artists — lines, bars, text, ticks. Two ways to drive it:

• pyplot (state machine): plt.plot(x, y) — quick, implicit "current axes".
• Object-oriented (explicit): fig, ax = plt.subplots(); ax.plot(x, y) — what you should use for anything you'll re-edit.

import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(8, 4))
ax.plot(months, revenue, marker='o')
ax.set_title('Monthly Revenue')
ax.set_xlabel('Month'); ax.set_ylabel('₹')
fig.tight_layout(); plt.show()

The Five Charts That Cover Most Reports

Chart	Use it for
Line	trend over time
Bar	compare categories
Histogram	distribution of one numeric column
Box / violin	distribution across groups
Scatter	relationship between two numeric columns

Master these before pies, radar, or 3D anything.

Beginner Mistakes to Skip

1. Pie charts with 7+ slices. Humans can't compare angles. Use a sorted bar chart instead. 2. Dual-axis charts to fake correlation. Almost always misleads — separate panels are clearer. 3. Truncated y-axes that exaggerate tiny changes. Default to starting bars at zero. 4. 3D charts. Cool-looking, hard to read. Skip them. 5. Red/green only. ~8% of men are red-green colour-blind. Use viridis or cividis. 6. No title, no axis labels, no units. A chart without context is decoration.

Intermediate: Subplots & Layout

fig, axes = plt.subplots(2, 2, figsize=(10, 6), sharex=True)
axes[0, 0].plot(...)
axes[0, 1].bar(...)
axes[1, 0].hist(...)
axes[1, 1].scatter(...)
fig.suptitle('Quarterly snapshot')
fig.tight_layout()

sharex/sharey line up axes so the eye can compare across panels. tight_layout (or constrained_layout=True) prevents overlap.

Intermediate: Seaborn — Stats Built In

import seaborn as sns
sns.set_theme(style='whitegrid')
sns.barplot(data=df, x='country', y='revenue', errorbar=('ci', 95))
sns.boxplot(data=df, x='category', y='price')
sns.histplot(df, x='age', hue='gender', kde=True)
sns.scatterplot(data=df, x='ad_spend', y='revenue', hue='channel')
sns.heatmap(df.corr(numeric_only=True), annot=True, cmap='coolwarm', center=0)

Seaborn loves long-format DataFrames (one row per observation). It auto-computes 95% CIs on bars, KDEs on histograms, regression lines on regplot.

Intermediate: Styling That Looks Professional

plt.rcParams.update({
    'figure.dpi': 110,
    'font.size': 11,
    'axes.spines.top': False,
    'axes.spines.right': False,
    'axes.grid': True,
    'grid.alpha': 0.3,
})

Drop top/right spines, soften the grid, bump DPI for crisp screenshots. Pick one font, one accent colour, stick to it across the report.

Intermediate: Annotations Tell the Story

ax.axhline(target, color='gray', linestyle='--', label='Target')
ax.annotate('Pricing change',
            xy=('2026-03', 1200), xytext=('2026-01', 1500),
            arrowprops={'arrowstyle': '->'})
ax.legend(frameon=False, loc='upper left')

A chart with a callout converting raw lines into a *story* always beats a clean unlabelled chart.

Advanced: Choosing the Right Colour Palette

• Sequential (Blues, viridis) — ordered low→high (e.g. revenue heat).
• Diverging (coolwarm, RdBu) — meaningful midpoint (profit vs loss, change vs baseline).
• Categorical (Set2, tab10) — distinct unordered groups; cap at ~7 categories.

Always check with a colour-blind simulator (viridis and cividis are perceptually uniform and CB-safe).

Advanced: Performance for Large Data

• ax.scatter(x, y, alpha=0.1, rasterized=True) — rasterise heavy point clouds; lines/text stay vector.
• For >1M points: datashader or plotly + WebGL — they bin pixels server-side.
• Save final figures with fig.savefig('out.png', dpi=200, bbox_inches='tight') for slides; .pdf or .svg for print.

Advanced: When to Switch to Plotly

Matplotlib/Seaborn = static reports. Reach for Plotly when you need:

• Hover tooltips (executives love them).
• Range sliders and zoom.
• Embedding in Streamlit / Dash / web apps.

The DataFrame stays the same; just import plotly.express as px and call px.line(df, x=..., y=..., color=...).

Advanced: Reproducible Figure Pipelines

Professional reports build figures in functions, not notebooks of copy-pasted cells:

def plot_revenue(df, ax=None):
    ax = ax or plt.gca()
    ax.plot(df['month'], df['revenue'], marker='o')
    ax.set_title('Revenue')
    return ax

Now the same function works in notebooks, dashboards, and CI screenshot tests.

Practice Path

1. Take any CSV, plot one of each: line over time, bar by category, histogram, boxplot, scatter — all with titles, labels, units. 2. Build a 2×2 subplot "snapshot" page with shared x-axis and tight_layout. 3. Replicate one of your matplotlib charts in Seaborn using long-format data. 4. Add an annotation + horizontal target line that turns the chart into a one-glance story.