Speedometer charts: love or hate

Pie chart hate is tired. In this post, I explain my speedometer hate. (Also called gauges,  dials)

Next to pie charts, speedometers are perhaps the second most beloved chart species found on business dashboards. Here is a typical example:

 

For this post, I found one on Reuters about natural gas in Europe. (Thanks to long-time contributor Antonio R. for the tip.)

The reason for my dislike is the inefficiency of this chart form. In classic Tufte-speak, the speedometer chart has a very poor data-to-ink ratio. The entire chart above contains just one datum (73%). Most of the ink are spilled over non-data things.

This single number has a large entourage:

- the curved axis
- ticks on the axis
- labels on the scale
- the dial
- the color segments
- the reference level "EU target"

These are not mere decorations. Taking these elements away makes it harder to understand what's on the chart.

Here is the chart without the curved axis:

Here is the chart without axis labels:

Here is the chart without ticks:

When the tick labels are present, the chart still functions.

Here is the chart without the dial:

The datum is redundantly encoded in the color segments of the "axis".

Here is the chart without the dial or the color segments:

If you find yourself stealing a peek at the chart title below, you're not alone.

All versions except one increases our cognitive load. This means the entourage is largely necessary if one encodes the single number in a speedometer chart.

The problem with the entourage is that readers may resort to reading the text rather than the chart.

***

The following is a minimalist version of the Reuters chart:

I removed the axis labels and the color segments. The number 73% is shown using the dial angle.

The next chart adds back the secondary message about the EU target, as an axis label, and uses color segments to show the 73% number.

Like pie charts, there are limited situations in which speedometer charts are acceptable. But most of the ones we see out there are just not right.

***

One acceptable situation is to illustrate percentages or proportions, which is what the EU gas chart does. Of course, in that situation, one can alo use a pie chart without shame.

For illustrating proportions, I prefer to use a full semicircle, instead of the circular sector of arbitrary angle as Reuters did. The semicircle lends itself to easy marks of 25%, 50%, 75%, etc, eliminating the need to print those tick labels.

***

One use case to avoid is numeric data.

Take the regional sales chart pulled randomly from a Web search above:

These charts are completely useless without the axis labels.

Besides, because the span of the axis isn't 0% to 100%, every tick mark must be labelled with the numeric value. That's a lot of extra ink used to display a single value!

Comments

[David R Clark]

Perhaps the speedometer graphic is better put in the "visual metaphor" category rather than strictly considered a data visualization.
But I share your reaction.

[Cody Custis]

It took me awhile to figure out why I hate the speedometer chart so much.

The first is that a speedometer is used to measure a metric (vehicle speed) where one has a very specific goal and smaller deviance from that goal is better. Both a vehicle going 45 mph in a 70 mph zone or one going 100 mph in the same zone are (depending on locality) engaging in illegal behavior. Likewise, when going in a 25 mph zone, the goal is very different. In the European gas example (and especially in the regional sales), "higher is better" looks to be the case, meaning the speedometer chart is inappropriate.

The second is that the circular nature of the speedometer chart trades a more complicated angle comparison in exchange for using less space than a linear chart (numbers arranged linearly are pi times as long as numbers arranged in a circle). This makes sense when looking a deviance from a goal, it doesn't make sense when the goal is to have a strict measure.

[John Allison]

Speedometer chat will be good for visualisation of child mortality rate. Where higher values indicated danger

[Kaz]

It's worth pointing out that gauge charts have something in common with bullet charts: they can show context in addition to the single value. While you only see one value, the other elements give you additional information like "how close to the target is it" (target indicator), "what are the maximum and minimum values we expect to see" (what values are on the gauge), and "are any of these values considered good or bad" (coloring of the gauge).

It is still inefficient, but it may be better to think of a gauge chart as a distorted bullet chart, which explains one reason why they are still popular.