A chart without annotation is like a photograph without a caption. The viewer sees something, interprets it through their own lens, and often gets it wrong. Annotation is the bridge between raw data and intended meaning.
The word matters. Annotation is not decoration. It is not a title slapped on top or a subtitle shoved underneath. It is a precisely placed piece of text that tells the reader what to look at, what it means, and why it matters. The best annotations are invisible in the sense that the reader does not notice them as separate from the chart. They are part of the reading experience.
What to Annotate
Not everything. The most common mistake is to annotate every data point, creating a chart that is more text than graphic. Effective annotation is selective. It draws attention to what is noteworthy and lets the rest of the data speak for itself.
Annotate anomalies. A sudden spike, a sharp drop, an unexpected plateau — these are the moments where the reader will ask "why?" and the annotation should answer. "Server outage, 3 hours" next to a traffic dip. "Product launch" next to a revenue jump. "Holiday weekend" next to a volume trough.
Annotate context. If a chart shows quarterly earnings, annotate the quarter when a major acquisition closed. If a chart shows user growth, annotate the month when a competitor shut down. Context transforms data from a pattern into a story.
Annotate the key finding. If the chart exists to make a specific point, state that point directly on the chart. "Revenue growth slowed in Q3 for the first time in two years." Place this text where the reader's eye naturally goes — near the inflection point, not in a distant corner.
How to Annotate
Position annotations close to the data they reference. A line connecting an annotation to a distant data point is a sign that the annotation is in the wrong place. Move it closer. Reduce the connector to a short tick mark or eliminate it entirely.
Use a smaller font size than the main text. Annotations are supporting actors. They should be legible but subordinate to the data. A sans-serif font at 11 or 12 pixels, in a muted color, works well against serif body text.
Keep annotations short. One sentence is ideal. Two sentences are the maximum. If the explanation requires a paragraph, it belongs in the body text, not on the chart.
The New York Times Standard
The graphics desk at The New York Times sets the standard for annotation in data journalism. Their charts routinely include three to five annotations that guide the reader through the data in a specific sequence. The annotations are not afterthoughts — they are designed simultaneously with the chart, shaping the visual hierarchy.
Amanda Cox, formerly of The Times, described charts as "the most efficient way to tell a story with data." Annotations are what make the chart a story rather than a display. Without them, the reader sees shapes. With them, the reader sees meaning.
Annotation in Dashboards
Dashboards rarely include annotations. This is a significant failure. A dashboard metric that changed by 15% last week begs the question "why?" — and the dashboard almost never answers. The reader must leave the dashboard, open a different tool, and investigate independently.
Adding annotation capabilities to dashboards — even simple text notes attached to time periods — would transform them from monitoring tools into understanding tools. Some platforms, including Tableau and Observable, support this. Most business intelligence tools do not prioritize it.
The argument against annotation in dashboards is that dashboards should be "objective" — just the numbers. This misunderstands objectivity. A number without context is not objective. It is incomplete. Saying "churn increased 12%" without noting "we raised prices 20% last month" is not objectivity. It is omission.
The Craft
Annotation is a writing skill as much as a design skill. The annotator must decide what is important, express it concisely, and place it where it will be read at the right moment. This requires understanding the data, understanding the audience, and understanding the visual flow of the chart.
It is, in short, editorial work. And like all editorial work, it improves the thing it touches.