By Michela Tjan
To learn more about the Open Visualization Academy (OVA), visit https://openvisualizationacademy.org
Open almost any chart and you'll find everything except the point.
The axes are labeled. The colors are tasteful. The data is clean. And still you sit there for a second or two, eyes drifting across the bars, waiting for someone to tell you what you're supposed to notice. Nobody does. The chart is finished, technically. It just isn't saying anything.
That gap between a chart that's complete and a chart that communicates is almost always the annotation layer. And annotation is almost always the thing we add last, if we add it at all.
An example of a data viz that’s complete vs. communicates
Annotation isn't decoration
We tend to treat annotation like garnish. A title here, a callout there, an arrow if we're feeling generous. Cleanup work for the end of the project, after the "real" visualization is done.
That framing is backwards.
The data layer shows what happened. The annotation layer says what it means. One is description; the other is argument. The moment you circle a single point and write this is when we changed the pricing, you've stopped presenting data and started making a claim — and a claim is the only thing a reader can actually do something with.
What the layer is actually made of
It helps to stop treating annotation as one thing. In practice it's several, each doing a different job:
Orientation — titles, units, axis labels. The basics that let a reader enter the chart without a decoder ring.
Direction — a line that tells them where to look first. "Read down the left column" isn't hand-holding; it's respect for someone who has six seconds.
Emphasis — the highlighted line, the muted background series, the single colored bar in a field of gray. This is you deciding, on the reader's behalf, what matters.
Explanation — the callout that names the cause. The spike isn't interesting until you say it's the week the policy changed.
Four jobs the annotation layer does: 1. Orientation 2. Direction 3. Emphasis 4. Explanation
Notice that only the first of those is neutral. The rest are editorial. You're choosing what to surface and what to let fade. That choosing is the work.
The test I keep coming back to
When a chart feels flat, I've stopped asking is this accurate? and started asking something harder: if I covered the data and left only the words, would a reader still know what I’m trying to convey?
If the answer is no, the chart is doing all the lifting and the annotation is doing none of it. That isn't clean, minimal design. It's an unfinished argument wearing minimalism as a disguise.
The fix is rarely more data. It's usually one sentence, placed where the eye already wants to land.
Restraint, still
None of this is permission to bury a chart in sticky notes. The strongest annotation is sparse — it earns its place by removing the reader's need to guess, not by narrating every tick. Tools like Datawrapper and D3 make it trivial to drop text anywhere on a canvas, which is exactly why the discipline has to come from you and not the toolbar.
A rule I trust: every annotation should answer a question the reader was already about to ask. If it doesn't, it's clutter in a helpful costume.
Annotation as clutter vs purposeful
In other words, a chart without annotation hands the reader a map and walks off. A chart with it points and says, here → this is the part that changes your mind.
