Data Visualization

Data visualization is the graphical representation of data and information. The human visual system detects patterns, trends and outliers in a picture far faster than in a table of numbers — which is why visualization is both an analysis tool (EDA) and a communication tool (reporting).

Why Visualize?

Choosing the Right Chart

Chart Reference Table

Principles of Effective Visualization

1. Edward Tufte's Data-Ink Ratio

                     Ink used to display data
   Data-Ink Ratio = ───────────────────────────      ->  maximise this
                       Total ink used in the graphic

Remove chartjunk: 3-D effects, heavy gridlines, decorative backgrounds, unnecessary borders, gradient fills, drop shadows.

2. Core Rules

3. Common Misleading Practices

Visual Encoding — Ranked by Human Accuracy

Cleveland and McGill's classic ranking of how accurately people decode visual channels:

1. Position along a common scale     <- most accurate (bar, scatter)
2. Position on identical, non-aligned scales
3. Length
4. Angle / Slope                      (pie charts live here — hence their weakness)
5. Area                               (bubble charts)
6. Volume / Depth
7. Colour saturation / Shading        <- least accurate

Design implication: whenever precision matters, encode with position or length, not with area or colour.

Python Visualization Ecosystem

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

sns.set_theme(style="whitegrid")     # clean default styling

sales = pd.DataFrame({
    "month":  ["Jan", "Feb", "Mar", "Apr", "May", "Jun"],
    "revenue": [420, 465, 448, 512, 590, 634],
    "region": ["North", "North", "South", "South", "East", "East"],
})

fig, axes = plt.subplots(2, 2, figsize=(13, 9))

# 1. LINE — trend over time
axes[0, 0].plot(sales["month"], sales["revenue"], marker="o", color="#168B99", linewidth=2)
axes[0, 0].set_title("Revenue Grew 51% from Jan to Jun", fontsize=12, fontweight="bold")
axes[0, 0].set_ylabel("Revenue (Rs lakh)")

# 2. BAR — comparison, sorted, starting at zero
axes[0, 1].bar(sales["month"], sales["revenue"], color="#10b981")
axes[0, 1].set_ylim(0, 700)                       # zero baseline — honest comparison
axes[0, 1].set_title("Monthly Revenue")

# 3. PIE — composition (few slices only)
region_totals = sales.groupby("region")["revenue"].sum()
axes[1, 0].pie(region_totals, labels=region_totals.index, autopct="%1.1f%%",
               colors=["#168B99", "#10b981", "#f59e0b"])
axes[1, 0].set_title("Revenue Share by Region")

# 4. HORIZONTAL BAR — long labels read better horizontally
axes[1, 1].barh(sales["month"], sales["revenue"], color="#6366f1")
axes[1, 1].set_title("Revenue by Month")
axes[1, 1].set_xlabel("Revenue (Rs lakh)")

plt.tight_layout()
plt.show()

# Removing chartjunk — before/after
fig, ax = plt.subplots(figsize=(7, 4))
ax.bar(sales["month"], sales["revenue"], color="#168B99")

# Maximise data-ink: strip the top/right spines and heavy gridlines
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.grid(axis="y", alpha=0.3)
ax.set_axisbelow(True)
ax.set_title("Revenue climbed every month except March", loc="left",
             fontsize=13, fontweight="bold")
ax.set_ylabel("Rs lakh")

# Direct labels beat a legend
for i, v in enumerate(sales["revenue"]):
    ax.text(i, v + 8, str(v), ha="center", fontsize=9)

plt.tight_layout()
plt.show()

Dashboards

A dashboard presents multiple related visualizations on one screen for monitoring key metrics.

Design rules: most important metric top-left (F-pattern reading), 5–9 visuals maximum, consistent colour meanings, filters/slicers for interactivity, and every chart must answer a question someone actually asks.

The next three lessons take the three most examined chart types — histogram, box plot, scatter plot — and go deep on how to build and read each one.