Tableau makes dashboarding a very fluid and creative process. The possibilities for displaying data and the ways in which you can drill down and filter are endless. This is great, but you must take care to choose wisely among the many options available, or your audience may become data-saturated.
To prevent this, we need to provide different levels of detail based on who is consuming the data and the type of task they are performing. I like to categorize dashboards into the following categories, in increasing level of detail: high-level, exploratory, and task-level. For today’s post we will focus on high-level dashboards.
Less Can Be More
On your car’s dashboard, you can easily read the speedometer and fuel gauge. When your turn signal is on, a light flashes, and if something requires your attention, a warning indicator lights up. Although car manufacturers could easily add another dozen gauges to tell the driver “more,” it would reduce the effectiveness of the display.
In the same way that you need only the essential information when driving, many data consumers at the management level need to see concise summaries without exploring the underlying data. A good high-level dashboard is:
- Read-Only: It could be printed or displayed on a wall monitor and would still fulfill its purpose.
- Bold and Sparse: Fonts are large. Color and icons may be used to emphasize whether metrics are within expected ranges, improving, or degrading.
- Actionable: Every metric shown has the potential to drive the user to ask further questions or take action if the value is outside of the normal range.
Let’s See an Example
The dashboard below, created by my Marquis associate G. Gabriel Kluepfel, meets some of the requirements but also has some room for refinement to be an effective high-level view:
As you can see, it’s very easy to glance at the map and see what’s going on by state. Clear sales numbers and color-coding by profit margin make it obvious which states are performing well. Since this does not give a sense for whether the situation is improving or degrading, Gabriel provided additional panels that show the trend at a glance.
However, there are some design elements that aren’t a perfect fit for a high-level dashboard. For one thing, it shows state-level data alongside item-level data. People who are responsible for sales and margins by state may not be interested in the individual items. This dashboard is also designed to be interactive, with selectors for individual states and items. My guideline for high-level dashboards is that they should show what the users want to see without interaction.
To adjust this to be a high-level dashboard, I’ll first remove all parameter adjustments and replace them with fixed filters that are clearly indicated. Next, I’ll remove item-level details so my viewer can focus on the big picture. Finally, I’ll determine which metrics are absolutely necessary to understand the situation. For example, do I need to include revenue, cost, and margin, or would revenue and margin be enough?
After making these changes, the dashboard is much cleaner and focused:
I kept the elements that may drive a manager to ask questions without adding details that may be a distraction. From here, step back – literally a couple of feet from your display. From a distance, you will find that some of the text is too small to read. Besides making fonts larger, you can see Jared’s post last week for techniques to reduce clutter and improve readability.
Stay tuned for upcoming posts that show how to build this paneled look and how to build dashboards for “what-if” analysis and exploration.
Chris Beck is a Principal Consultant for BI Solutions at Marquis Data.