Key Takeaways — Chapter 7: Typography, Annotation, and the Words on Your Chart

1. Decluttered Does Not Mean Self-Explanatory

Chapter 6 showed how to clean visual noise from a chart. This chapter showed that cleaning is not enough — a clean but wordless chart is readable but not understandable. The reader needs words to know what the data is, what the finding is, what the units are, and where the data came from. Decluttering and adding words are both necessary, and they are separate steps: first remove what is in the way, then add what is essential.

2. The Action Title States the Finding

The threshold concept: an effective chart title tells the reader what to conclude, not what to look at. A descriptive title ("Quarterly Revenue") names the topic. An action title ("Revenue Grew 18% in 2024") states the finding. For most real-world communication contexts — business reports, news graphics, dashboards, social media — the action title is better because it enables the reader to grasp the main message in the 5-second reading window. Writing an action title is a disciplined act of interpretation and is probably the single highest-impact change most practitioners can make to their chart output.

3. Every Chart Has a 5-Second Test

A self-explanatory chart should be understandable within five seconds by a viewer who has never seen it before. The test asks: can the viewer state what the chart is about, what the main finding is, and what the units are, in a single glance? Most charts fail this test, and the failure is almost always in the words. The action title, the subtitle, the axis labels, and the annotations are what make the test passable.

4. Typography Is a Craft of Restraint

Five principles govern good chart typography: use a single legible sans-serif font family, establish a clear size hierarchy, use weight (not color) for emphasis, align text meaningfully with the plotting area, and leave whitespace around every piece of text. None of these principles is controversial. All of them are ignored by default plotting libraries. Applying them deliberately is one of the fastest ways to make your charts look professional without adding any data or content.

5. Annotations Are the Text That Does the Most Work

An annotation is a short piece of text placed directly on the data, calling out a specific feature and explaining its meaning. One to three annotations per chart is the usual range. Each annotation should be short, specific, and attached to the feature it describes. Annotations are underused in default plotting because the libraries do not make them easy, but the cost of learning to use them is repaid many times over in reader comprehension. A single well-chosen annotation can carry the weight of a paragraph of explanation.

6. Axis Labels Need Units, and Tick Labels Need Formatting

Default axis labels almost always omit units, and default tick labels almost always show raw numbers without thousands separators, natural-unit scaling, or human-readable date formats. Specific fixes: put the unit in the axis label inline ("Revenue (USD millions)"), scale tick labels to natural units, use thousands separators, reduce the tick density to five or six major ticks per axis, and format dates for human reading. These small changes turn default axes into professional ones without any design expertise.

7. Direct Labeling Usually Beats a Legend

When a chart has multiple categorical series, direct labeling places the series name next to its line or bar, eliminating the eye movement between the data and a separate legend box. Direct labels are faster to read, survive cropping when the chart is shared, and reduce the color burden because the reader does not need to distinguish all colors — only match each label to its nearest line. Legends are appropriate for many-series charts, crossing lines, and interactive contexts, but for most charts, direct labels are the better default.

8. Source Attribution Is Non-Negotiable

Every published chart should include an on-image source attribution: a single line at the bottom of the chart naming the data provider, the vintage, and any essential methodology notes. Attribution must be on the chart itself, not in a caption, because captions get stripped when charts are shared, screenshotted, or embedded. Source attribution is a trust mechanism, not just a citation convention: it gives the reader a path to verification and makes the chart feel credible. Build attribution into your default workflow so it appears automatically on every chart you produce.

9. Context Notes Are a Form of Honest Communication

When the data has a wrinkle the reader needs to know about — a methodology change, an excluded category, an adjustment — include a short context note near the source attribution. "Methodology updated in 2018." "Revenue in constant 2020 dollars." "Excludes Alaska and Hawaii." These notes are not apologies; they are transparency statements. A chart that includes them is doing the work of responsible communication; a chart that omits them lets the reader form impressions that the full data would not support.

10. The Discipline Is Cheap and Compounds Across Charts

Writing an action title takes under a minute. Formatting tick labels is a few lines of matplotlib code. Adding a source attribution is a one-line fig.text call. Placing an annotation is a one-line ax.annotate call. The per-chart cost of applying every principle in this chapter is small, and the return in reader comprehension is large. Across a body of work — a report, a dashboard, a portfolio of published charts — the discipline compounds into a recognizable personal or institutional style that makes your charts readable, trustworthy, and memorable.