The Use (and Misuse) of Gauges

by Kirk Cunningham

Using a gauge as a graphical element for representing data can be very effective when used properly, but has some potential pitfalls that a data visualizer will want to avoid. In addition to the general caution not to take the term "dashboard" too literally, some specific concerns with the use (or misuse) of gauges center on:

• They often convey too little information.
  (All that artwork just to present one data value?)
• Many times, there is no context.
  (A needle points to $2,345 revenue. How far is it from goal? How does that compare to last month?)
• They can take up too much space.
  (They're as big as a speedometer instead of a temperature gauge.)

To avoid such pitfalls, consider these best practices for using gauges in your dashboards - based on some basic data visualization principles:

Use a gauge to represent a dynamic data value. You always want to use the graphical element that will give the best insight into your data; a bar chart is effective to show trends; a pie chart to show proportions. Gauges best represent variables that are readily changed or changing.

Best Practice: Since a gauge generally points to one value, the key is interactivity. Use a gauge to represent a data value that changes with different activities or scenarios (e.g. just-in-time inventories, KPIs, pro-forma revenues and expenses, variances); anything you're tracking or projecting. It's best used to represent a target objective that will vary over time (if tracking current activity) or vary with different assumptions (if considering what-if scenarios). In other words, use a gauge to represent the result of real-time or projected changes in the underlying data. And, remember to include context!

Context is essential. A data value, by itself, conveys little useful information. To be useful (and actionable) a data graphic must answer the question, "compared to what?"

Best Practice: When you use a gauge, as with any graphical element, always provide context. This can be accomplished in many ways, including:

• Always include alerts. Alerts are the visual zones on the gauge that indicate when the data value (needle) is too far above or below target. These will indicate at a glance if the underlying variable requires corrective action.
• Always label the endpoints. The hands of a clock are useful because you know what their position and the distance between them mean. Similarly, it should be apparent what the position of the gauge needle means, and how much action is required to change that position.
• Add context by including additional graphical elements. You can also include other related data in your data presentation or dashboard. For example, a gauge doesn't show trends. So, consider adding a small bar chart to your dashboard that gives trend data. The viewer will quickly see if the underlying variable is on target, what the variance is, and how that has changed over time.

Don't confuse data visualization with art. Many data presenters take their bosses too literally when asked for a Corporate dashboard. Sure, a Ferrari speedometer looks sexy, but if designed properly, a small simple gauge can be far more effective...especially when used as small multiples.

Best Practice: A better analog than a car's speedometer is its temperature gauge. It is small, unobtrusive and tells you at a glance the status, context and variance. For example, consider the use of a small gauge (or multiple small gauges) where "on plan" is with the needle straight up. To the right is a green-alert zone, indicating a positive variance. To the left is a red-alert zone indicating a negative variance. The endpoints are labeled "-10%" and "+10%" respectively. Now, with no detail needed, immediately the viewer will know if they are under/over performing on a KPI, and by how much. Then they'll look elsewhere on the dashboard for additional data to understand why they're not on target. Or, in the case of what-if projections, they'll run another scenario to try to achieve the objective.

Despite some recent bad press, a gauge isn't inherently a poor graphic. Like all other graphical representations of data (pie charts, radar charts, scatterplots, etc.) they must be employed in the appropriate situation, and they must follow the basic principles of good data visualization.