Visualising data honestly

Why charts matter

Most people who use your analysis will never read your code or your tables. A clear chart is how your work reaches managers, clients and the public. A misleading chart, even an accidental one, can lead to bad decisions. So choosing and drawing charts well is part of being a data professional, not decoration.

Pick the chart by the question

Drawing a chart with matplotlib

import matplotlib.pyplot as plt

cities = ["Pune", "Delhi", "Mumbai"]
sales = [1230, 1350, 2100]

fig, ax = plt.subplots(figsize=(6, 4))
ax.bar(cities, sales, color="#2563eb")
ax.set_title("Sales by city, March 2025")
ax.set_ylabel("Sales (₹)")
ax.set_ylim(0, None)          # bars must start at zero
plt.tight_layout()
plt.show()

Rules for honest charts

• Start the value axis of a bar chart at zero. Bars show size by length; cutting the axis makes a 2% difference look like 200%.
• Line charts may zoom in on a range, but say so clearly on the axis.
• Label axes with units (₹, minutes, %), add a title that says what and when, and name the data source.
• Do not use 3-D effects; they distort sizes.
• Keep the same scale when you compare charts side by side.
• Show the number of data points when it is small, so nobody over-reads 5 values.

Colour

About 1 in 12 men and 1 in 200 women have some colour vision deficiency, most often trouble telling red from green. Use colour-blind-friendly palettes (for example the "viridis" colour map or the Okabe-Ito colours), do not rely on red vs green alone, and use colour only when it carries meaning. One highlight colour against grey is often clearer than a rainbow.

Histograms and bins

A histogram splits a number range into equal bins (for example 0–10, 10–20 minutes) and draws how many values fall in each. The bin width changes the story: too wide hides detail, too narrow looks noisy. Try a few widths before choosing.

Common mistakes

• Pie charts with many slices. Fix: use a sorted bar chart.
• Two different y-axes on one chart that suggest a link that is not there. Fix: use two separate charts.
• Default titles like "Figure 1". Fix: write the takeaway: "Mumbai sold the most in March".
• Plotting averages without showing spread. Fix: add a box plot or error bars where it matters.