Visualization and Charts

Numbers Need Pictures

A table of 500 rows tells you nothing at a glance. A line chart showing the same data tells you everything in two seconds. Visualization isn't decoration — it's how humans actually understand data.

AI excels here because it can look at your data, recommend the right chart type, and generate the code to create it. You don't need to be a designer or a developer.

Picking the Right Chart

The wrong chart type can actively mislead. Here's when to use what:

Line charts: Trends over time. Monthly revenue, daily active users, temperature changes.

Bar charts: Comparing categories. Sales by region, votes by candidate, budget by department.

Pie charts: Parts of a whole — but only with 5 or fewer slices. More than that, use a bar chart instead.

Scatter plots: Relationships between two variables. Does ad spend correlate with conversions?

Heatmaps: Patterns across two dimensions. Website traffic by day and hour, for example.

Don't worry about memorizing this. AI will recommend the right type if you describe what you're trying to show.

Matching Data to Chart Type

Choosing the wrong chart is one of the most common visualization mistakes. Here is a decision framework you can use every time:

Common mistakes to avoid:

Using a pie chart with 12 slices — the human eye cannot reliably compare angles beyond 5 segments. Switch to a horizontal bar chart.

Using a line chart for unrelated categories — lines imply continuity and connection between points. If there is no natural order, use bars instead.

Using 3D effects — they distort perception and make data harder to read. Always use flat, 2D charts for accuracy.

Truncating the y-axis — starting at a number other than zero makes small changes look dramatic. If you must truncate, label it clearly.

Visual Design That Works

Good chart design follows a few universal principles. Ask AI to apply these when generating visualizations:

Data-ink ratio: Every pixel should communicate data. Remove gridlines, borders, backgrounds, and decorations that do not carry information. Less ink, more insight.

Color with purpose: Use color to highlight the important thing, not to decorate. One accent color for the key data point, muted tones for everything else. Never use red and green together — colorblind viewers cannot distinguish them.

Readable labels: Every axis needs a label. Every label needs units. If a viewer has to guess what the numbers mean, the chart has failed.

Annotation over decoration: Instead of adding clip art or fancy backgrounds, add annotations that point out the key insight directly on the chart. A callout arrow saying "Campaign launched here" is worth more than any gradient.

Consistent scales: When comparing two charts side by side, use the same y-axis scale. Different scales create the illusion of different magnitudes when the data may be similar.

The AI Visualization Workflow

Claude's analysis tool can generate charts directly in the conversation. Upload a CSV and ask for a visualization — you'll get an interactive chart right there.

import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv("sales_data.csv")
df["date"] = pd.to_datetime(df["date"])
monthly = df.groupby(df["date"].dt.to_period("M"))["revenue"].sum()

# Line chart with insight-based title
fig, ax = plt.subplots(figsize=(10, 5))
monthly.plot(kind="line", ax=ax, marker="o", color="#7c3aed")
ax.set_title("Revenue grew 34% after the March campaign launch", fontsize=14, fontweight="bold")
ax.set_ylabel("Revenue ($)")
ax.set_xlabel("")
ax.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.savefig("revenue_trend.png", dpi=150)
plt.show()

# Bar chart comparing categories
by_product = df.groupby("product")["revenue"].sum().sort_values()
fig, ax = plt.subplots(figsize=(8, 5))
by_product.plot(kind="barh", ax=ax, color="#f97316")
ax.set_title("Product C drives 42% of total revenue", fontsize=14, fontweight="bold")
ax.set_xlabel("Total Revenue ($)")
plt.tight_layout()
plt.savefig("product_comparison.png", dpi=150)
plt.show()