Data viz that doesn't look like a 90s Excel chart

Two libraries cover 95% of what you'll ever need:

• matplotlib — the foundation. Explicit, controllable.
• seaborn — sits on top of matplotlib, defaults are prettier, statistical charts one-liner.

matplotlib mental model

fig, ax = plt.subplots()     # figure = canvas, ax = one plot
ax.plot(x, y)
ax.set_title(...); ax.set_xlabel(...); ax.set_ylabel(...)
plt.tight_layout(); plt.savefig("out.png")

• plt.subplots(2, 3) — grid of plots; axes is a 2D array.
• Everything is customizable — colors, linewidth, markers, ticks, log scale.

When to use each chart

• Line — series over time, trends.
• Bar — categorical comparisons.
• Histogram — distribution of one variable.
• KDE (sns.kdeplot) — smoothed histogram.
• Boxplot / Violin — distribution across groups.
• Scatter — two variables, correlation.
• Heatmap — 2D matrix, e.g. correlation matrix.

seaborn one-liners

import seaborn as sns
sns.set_theme(style="whitegrid")
sns.barplot(data=df, x="region", y="revenue", hue="product")
sns.lineplot(data=df, x="date", y="sales", hue="channel")
sns.heatmap(df.corr(numeric_only=True), annot=True, cmap="RdBu_r")
sns.pairplot(df[["age","income","score"]])

Style rules

• Label everything — title, axes, units.
• Use color intentionally — brand color for the series in focus, grey for the rest.
• Never rely on color alone — colorblind-safe palettes (viridis, cividis).
• Sort bars by value, not alphabetical — makes ranking readable.
• Use log scale when values span orders of magnitude.

Where to next

• plotly for interactive HTML charts (great for reports + notebooks).
• altair for a grammar-of-graphics style declarative API.

Try it

1. Make a bar chart with the tallest bar in your brand color and the rest grey.
2. Overlay a rolling mean line on top of a daily scatter.