Improving data visualization: Where do I put all those annotations?
One of critical questions any improver hopes to be asked when presenting data is, “wow, those are really impressive results! How on earth did you achieve them?” The complete answer might be a long story, but often a few key parts of that story can be presented as time-series chart annotations.
Annotations are simply text that indicate a point on a time-series chart when a key event occurred. These key events help us understand the movement—up or down—in time-series data. Simple. What could possibly be complicated about that?
Well, sometimes annotated time-series charts can hard or confusing to read, especially when a chart has many of them. Here are some ideas about how to make annotations more clear.
The chart below isn’t “bad.” It includes good annotations and points out when they occurred, but it’s a little visually jarring and isn’t the best way to illustrate the story we want to tell. So how can we improve it?
Annotations should stand out, but less so than the data.
Good data visualization tells a story. The story we usually want to tell is: 1) we achieved these results and 2) here’s how we did it. Accordingly, we want to design a visualization that draws the audience’s attention first to the data (what we accomplished), and then second to the annotations (how we did it).
So how do we do this? First, people are naturally drawn to the biggest, brightest text or graphic on a slide. Second, people [in most languages around the world] read left to right. So, we want to make the data bright and big by using bold, darker colors and thick lines. I like 3.5 point lines for line charts in Excel and no more than 100% spacing between bars in bar charts in Excel. Conversely, make the boxes around the annotations lighter and less thick.
Annotations should not obscure the data.
This sounds obvious. Never let the data be covered up by annotations. Usually this can be accomplished by careful positioning, but sometimes we have so many annotations that in an effort to fit them all in, we end up covering parts of our time-series line chart.
One way to ensure that the annotations don’t obscure the data is to place them entirely off of the chart area. By placing them under the chart, and drawing lines to the dates on the chart x-axis, we can invoke the visual language of timelines, which are very familiar to most viewers.
I think these changes make for a clear, easy-to-read annotated chart that the viewer is prompted to read in the intended order. By trying the place the annotations directly below the month in which they occurred and drawing a vertical line from the month to the annotation, this method is usually better at ensuring annotations are read in chronological order than the chart above, where people will probably read the annotation for February first, or just be confused about where to start. This timeline of annotations won’t always work well—for example when there are many annotations in one or two months or just one or two total annotations. For one to four annotations, putting them on the chart area (as in the first example) probably works better.
Here are a few more ideas about improving annotations on times-series charts:
Annotations should be concise: Annotations should use as few words as possible to convey a clear idea. They don’t need to say everything. If there are more details that can be included while keeping the chart readable, leave the less important annotations to be said in an oral presentation or included in the main body text of a written report.
Lists of changes go to the left: Often, when presenting aggregated data from multiple health facilities, we want to present the change ideas that the facilities used to achieve their results. However, even when many or all facilities implemented the same change idea, they don’t always do so in the same month. As a result, we often include a boxed list of key change ideas implemented by the facilities without including the dates for specific changes. As mentioned above, because most viewers will read left to right, the boxed list of annotations should go on the right side of the chart so that they view the data first.
Use color to convey meaning: The color of the box around annotations can be used to distinguish different categories of annotations. In the example below, different colors for the boxes around the annotations are linked to the colors of the affected indicators on the time-series chart (green for maternal care and grey-scale for essential newborn care).
- Improving data visualization: Using overlapping bar charts to present actual vs target data
- Improving data visualization: No more excel data tables!