What makes a chart good

A chart is good when the reader understands the main point quickly and correctly. Not when it looks impressive, uses many colours or shows every number you have. This simple definition leads to practical rules.

Start with the question

Before choosing a chart, write the question it should answer in one sentence:

• "Which product category grew the most this year?"
• "Do weekend riders take longer trips than weekday riders?"
• "Is there a relationship between delivery time and review score?"

The question tells you which comparison matters, and that tells you which chart to use. A chart without a question usually ends up showing everything and explaining nothing.

How people read charts

Research on graphical perception, starting with work by Cleveland and McGill in the 1980s, found that people judge some visual encodings more accurately than others. Roughly from most to least accurate:

1. Position along a common scale (bar ends, dots on an axis)
2. Length
3. Angle and slope
4. Area
5. Colour intensity

That is why bar charts and dot plots are so effective for comparisons, and why pie charts and bubble sizes are harder to read precisely. It does not mean pie charts are always wrong, but it is a good reason to prefer bars when the exact comparison matters.

Remove what does not help

Every element on a chart should earn its place. Heavy gridlines, borders, 3D effects, shadows and legends that could be direct labels all compete for attention. Removing them makes the data stand out. This is sometimes called improving the data to ink ratio, an idea popularised by Edward Tufte.

Tell the reader what to see

A title like "Sales by month" describes the chart. A title like "Sales grew every month except August" tells the reader the point. In exploratory work, descriptive titles are fine. In anything you share, a takeaway title saves the reader effort and makes your conclusion explicit, so it can be checked.

Honest charts

• Bar charts should start at zero, because bar length encodes the value.
• Line charts can use a narrower range when the change is what matters, but say so clearly.
• Do not cherry pick a date range to exaggerate a trend.

In the next lesson we set up the tools and look at the structure of a matplotlib chart.