Visualization Genres: Monitoring and Dashboards
We’ve talked about exploration — learning new things about your data; and we’ve talked about presentation — teaching other people things about data. The third major genre of visualization is monitoring and dashboards —learning about new data in real time.
(In the next entry, I’ll say a little about things that don’t fit well into these categories.)
I’ll use the words “monitoring” (as the task) and “dashboard” (as the tool that we use to carry out the taxsk) somewhat interchangably. A dashboard is a set of fixed visualizations connected to live data, with the goal of letting a user have a consistent view of a standing question — and so, to monitor a situation.
The dashboard curator chooses a set of visualizations, backed by data queries. When a user looks at the dashboard, they see a recently-updated view.
Paradoxically, dashboards are perhaps the least-loved genre in the visualization world, yet the most used. Michael Correll has described them as “rude”; visualization researchers often dismiss them as uninteresting. They don’t come with the thrill of discovery, as exploratory visualization does; and they don’t encourage novel renderings, like presentations do.
Four dashboard exemplars showing different use cases. From Sarikaya et al's What Do We Talk About When We Talk About Dashboards? (IEEE VIS 2019).
Sometimes, a dashboard is used to communicate updates to others, as in the COVID-19 dashboards that were a regular part of life in 2020 and 2021. Other times, dashboards are a way of pinning down a frequently-asked question: at Honeycomb, we often saw our customers create dashboards that allowed them to easily check whether their system was working well, or whether a known issue was recurring.
Addressing known issues seems to be the key to dashboards: they pin down past questions, more then they explore future ones. A decision-support dashboard shows the answer to past questions: “last time this happened, we wanted to know the answer to this question. What about this time?”.
A well-designed dashboard invites two next steps:
Letting the reader compare a value to a past state, or desired ranges. Not all dashboards are good at comparison: is “12,750” in the top-left of the social dashboard, above, growing or shrinking? Is it a good number or a bad one? The reader might know, but the visualization doesn’t seem to say. Gauges, like the operational dashboard above, let the reader compare to a desired normative value, but may not provide information .
Asking questions about why a particular change has happened, and further exploration. (“Why is this number high?”)
At their worst, dashboards are retreads: Honeycomb CTO Charity Majors talks about a dashboard as “an answer to some long-forgotten question”, “an invitation to pattern-match the past instead of interrogate the present.”
That’s our three major genres — next time, I’ll talk a little about things that don’t fit so well into the categories, and try to put these in a grid together.
Read more
A few years ago, I got together with some colleagues from Tableau and Microsoft got together to write an academic article, “What do we talk about Dashboards”, which tries to break down families of dashboard uses. Some of those authors followed up with “Heuristics for Supporting Cooperative Dashboard Design”, exploring ways to think about how to frame dashboards more usefully, using a conversational metaphor.
If you want a more tactical approach, Stephen Few’s “Information Dashboard Design” is a useful discussion of how to create practical dashboards.
And, of course — if you’ve got visualization challenges of your own, I’d love to help out! Drop me a line, and let’s talk about how to help our users and customers make sense of critical data.