Reading and Making Charts

Introduction

A good chart lets you see a pattern in seconds. A bad one can hide it, or invent one. Charts are used to persuade, so reading them critically is as important as making them.

Choosing a chart

Start with the question.

Question	Good chart
How do categories compare?	Bar chart
How does something change over time?	Line chart
How are values spread out?	Histogram
Is there a relationship between two numbers?	Scatter plot
What share does each part make up (few parts)?	Pie chart, or better, a bar chart

Notes:

A bar chart is the safest all-round choice for comparing amounts.
A line chart connects points in time order, so use it only when the order matters.
A histogram groups numbers into ranges (bins) and shows how many fall in each, revealing the shape of the data.
A scatter plot places each record as a dot by two values. A rising cloud of dots suggests that as one goes up, the other tends to as well.
Pie charts are hard to read once there are more than a few slices, because humans compare angles poorly.

Quick quiz

Which chart is best for showing how monthly sales changed over a year?

How charts mislead

A cut-off axis. If a bar chart starts at 90 instead of 0, a small difference looks huge. Bars should start at zero, because their length stands for the amount.
Stretched or squashed scales. Making a chart very tall or very wide changes how steep a trend looks.
Cherry-picked time ranges. A few months chosen to show a rise, while the longer picture shows a fall.
Missing context. No labels, no units, no source.
Too many colours or decorations, which distract from the data.
Dual axes, with two different scales on one chart, which can make unrelated lines appear linked.
3D effects, which distort sizes.

Always check the axes, the units, the time range and the source.

Making a clear, honest chart

Say what the chart shows in the title, ideally the main message.
Label both axes, with units.
Start bars at zero.
Keep it simple: few colours, no clutter.
Sort bars in a meaningful order, such as largest to smallest.
Use colour with purpose, and make sure it works for people who cannot tell some colours apart.
Add the source and the date of the data.
Show the number of records if it is small, because small samples can mislead.

Tools

You can make good charts with free spreadsheet programs. Later, programming libraries such as matplotlib in Python let you create charts from code, which you will do in the Data Analysis with pandas course.

Try it yourself

Find a chart in a news article or advert. Check its axes, units, time range and source. Write one sentence on what it really shows, and one on how it might mislead.

Summary

Choose the chart by the question: bars to compare, lines for time, histograms for spread and scatter plots for relationships.
Watch for cut-off axes, odd scales, cherry-picked ranges and missing context.
Honest charts have clear titles, labelled axes, zero-based bars and a stated source.