Should you use dashboards to tell stories or to find stories?
You can’t tell a good story if you don’t understand the data, and a good dashboard will help you understand the data, faster.
My thanks to Adam McCann, Athan Mavrantonis, Andy Cotgreave, and Jeffrey Shaffer for reviewing and providing feedback.
Why do we need dashboards?
Shortly before The Big Book of Dashboards was published, I had the good fortune to attend Cole Nussbaumer Knaflic’s Storytelling with Data workshop. Great content, great presentation, and I got… worried. I remember thinking “this is really good… why will anybody need the dashboard book?”
It was shortly after this momentary panic that Cole described a tendency that many people have in presenting their data. They want to relate just how long and hard they worked. “Look at all these data sources! Look at all the rows… look at all the columns! Look at every stone I unturned!”
Cole then related a wonderful metaphor:
“You have to shuck a lot of oysters to find a pearl. When presenting, don’t show all the oysters you shucked, just show the pearl.”
With these words I became calm as I realized what is so useful about dashboards.
Dashboards can be like automated oyster shuckers; that is, you can use them to find the best stories in your data and to find those stories—fast.
What you then do with those pearls… well that’s where the storytelling part comes in. Yes, you can tell stories in a dashboard (I’ve seen some amazing things) but I don’t think that’s the best use for them. Indeed, I think it’s fine for dashboards to be… boring.
Let me make my case.
Note: Are these “pearls” in fact stories, or just great insights from which you can craft a story? And just what is a “data story” ? I’ll address this towards the end of the article.
So, just what is a dashboard, anyway?
Those that follow my work or the work of my fellow book authors, Andy Cotgreave and Jeff Shaffer, know that the three of us agonized over the definition of a dashboard. After months we agreed on this intentionally open-ended definition:
A dashboard is a visual display of data used to monitor conditions and/or facilitate understanding.
Note we didn’t get into explanatory vs. exploratory, interactive vs. static, fits on one screen vs. scrolls vs. drills down to another screen, etc.
I know that Stephen Few took exception to our writing “and/or” (versus just “and”) and that Nick Desbarats might argue that our imprecise definition may be contributing to mismanaged expectations. I’ll come back to some of Desbarats’ concerns at the end of this post, but for the time being let’s just run with our definition and I’ll walk you through two examples where we see how a simple dashboard will facilitate understanding which in turn can lead to superior data visualizations and storytelling.
Example One – If the world were 100 people
This first example comes from Makeover Monday exercise from 2017. Here’s a snippet of the data
Figure 1 – see https://www.100people.org/statistics_100stats.php?section=statistics
And here was the visualization that people were tasked with making over.
Figure 2 – see https://visual.ly/community/infographic/geography/world-100-people
I wanted to get a solid handle on this data, so here’s the visualization I built.
Figure 3 – My rather uninspiring Makeover Monday submission.
I would never publish this. Okay, I *did* publish this, but would not expect anybody to act or be moved by anything I built. I made the dashboard simply to help me understand the data.
So, what might I do with it, now that I have a handle on it? What facet might I want to highlight and what story might I want to tell?
Maybe I want to focus on nutrition.
Figure 4 – Focusing on the 1 out of 100.
What does it mean that 1 out of 100 people is starving?
Here’s what Athan Mavrantonis did with this data.
Figure 5 – Athan Mavrantonis’ stark, and stunning, visualization.
This is a simple and elegant visualization, but also very powerful. Indeed, you don’t need to create something complex to have an impact.
This visualization is also beyond my design abilities. But, had I wanted to create something like this for a magazine I might have hired somebody with Mavrantonis’ skills—but only after I first had a solid understanding of the data.
Note that I have no idea if Mavrantonis started first with a simple bar chart, considered the facets of the data set, and then created this powerful image. But whether that was the case or not, I do know the following to be true:
You can’t tell a good story about the data if you don’t have a clear understanding of the data.
Example Two – Visualizing Eli Manning’s Career
I am a big fan of Adam McCann’s work and how he shares his expertise. I think of him as the Wynton Marsalis of data visualization as he is uncommonly adept with both straightforward business visualizations (“classical music”) and more experimental forays (“jazz”).
He recently published this interactive data visualization about New York Giant’s quarterback Eli Manning’s career.
Figure 6 – Adam McCann’s creative and engaging interactive dashboard.
Any time I see a graphic (let alone a novel and creative one like this) I go through the same thought process:
- What is this telling me?
- Why did this person choose this way to display the data?
- How did this person build this?
- Is this how I would have visualized the data?
Ut, oh… beware of those last two points as you’ll end up going down the rabbit hole (as I did).
Note that Adam’s dashboard was downloadable, so I could see how he built it and leverage whatever work he had done to create my own take on the data. Indeed, Adam had done a lot of the “oyster shucking” already with respect to setting up the data, preparing calculations, establishing a color palette, and so on.
Here’s what I came up with.
Figure 7 – My makeover of Adam McCann’s dashboards.
This is not visually arresting, and some might consider it boring.
But wait! There’s an amazing “pearl” in the data and it’s right there! Can’t you see it?
Well, of course you can’t.
How about now?
Figure 8 – The story I found, highlighted and annotated.
This finding makes the 2011 Super Bowl win even more impressive to me, and because I had a simple dashboard, I was able to find this pearl quickly.
So, now that I have a pearl, what should I do with it? This is where the storytelling part comes in as I might want to explore how the Giants managed to sneak into the playoffs and despite their underdog status won the three playoff games and then the Super Bowl.
You should ask yourself the same question of what you find in your data.
What you do with the curated results (the pearl) is up to you. Maybe you use Storypoints in Tableau, or some form of scrollytelling, or craft a tight presentation in PowerPoint (or, heaven forbid, create a longform dashboard) but realize that you can’t do anything until you have a solid understanding of the data.
And for me, *that’s* the best reason to use dashboards.
Postscript: I encourage you to watch Nick Desbarats’ presentation The 13 Types of Displays That Are All (Unfortunately) Called “Dashboards.”
In this presentation, Desbarats asserts that
- Most dashboards fail (I agree with this.)
- We need to have a clear taxonomy of dashboard types so we can manage audience expectations (I disagree with most of this, but you should still watch.)
- There’s a right way and wrong way to garner buy-in from your stakeholders (this comes around 70% into the presentation and is absolute gold.)
Just What is a Story, Anyway?
One of my book co-authors, Jeffrey Shaffer, asked me to read this article from Tableau Iron Viz champion Josh Smith. I now realize that how I use the word “story” and others use it may be wildly different. Smith points out that telling s story involved characters, a plot, a narrative, tension, release and so on. In my case I was equating “where’s the good story in the data?” with “what in the data is worth turning into a story?” These are two different things so I will clarify and say I think a dashboard is a great place to find what in the data is worth elevating into a story.
As for how you tell that story? I can recommend the following resources:
Storytelling with Data by Cole Nussbaumer Knaflic
Presentation Zen by Garr Reynolds
Slideology by Nancy Duarte
Postscript
I asked Athan Mavrantonis to review this post and share his thoughts on building his “If the world were 100 people” visualization. Here’s what he had to say:
I had not done any exploration before designing it. I created the initial visualization very quickly in Illustrator and only replicated it in Tableau later that day, upon Rodrigo Calloni’s request.
I was familiar with the dataset prior to MakeoverMonday. Bill Gates had used this data before and tweeted a stacked area chart showing the trends for each dimension over time from the 1970s to 2017. His point was very optimistic as the percentages of the people who were undernourished or starving were dropping significantly over that period. When I saw this viz, I remembered thinking his way of representing the data wasn’t fair because it ignored the exponential growth in the world’s population over the same period.
So, when I came across “The world as 100 people” dataset in that MakeoverMonday even as a point in time, I knew immediately what I wanted to visualize. I wanted to offer a different angle comparing the relative to the absolute numbers side by side. I chose to visualize only the number of people starving specifically because it was only 1 out of 100. I felt this chart would help the audience resonate with the one black circle in the context of it representing a unique, starving, human being. On the other side I presented a BAN to emphasize how many people were hidden behind that single dot.
I think my design and layout decisions are clear. Tried to keep it as simple and direct as possible, I didn’t use any color, and I focused on contrast between the two different perspectives.
Great post, Steve (and thanks for the shout out!).
I should clarify, though, that the main reason why I’m proposing to organize dashboards into a taxonomy isn’t to set audience expectations, it’s more so that we, as dashboard designers, can have less confusing conversations around dashboard design best practices, and to make life easier for people who are trying to learn when and when not to apply those best practices. I think that this need is illustrated in your post when you suggest that dashboards should be boring, but then, a little later, discuss creating a “non-boring” visual for communicating the insight that you’d found to an audience. It would be understandable if a novice designer read this and was left wondering whether they should make their dashboards boring or not.
I know that the answer to that question is Andy’s mantra (“it depends”), but I think that it’s worth trying to go a bit further to see if we can codify exactly WHAT it depends on, because “it depends” isn’t very helpful to people who don’t yet have a lot of dashboard design experience. I suspect that a good first step will be for us, as dashboard designers, to stop using the same term for, for example, a display that we use ourselves to spot new insights and a display for recommending a course of action to an audience, since the ways to go about designing those two types of displays effectively are almost completely different.
Now, whether “it depends” can be usefully captured in a taxonomy or any other type of codified framework is a very valid question. I suspect that it can but am not completely convinced. If people start pointing out tons of legitimate cases that fall outside of my (or someone else’s) framework and that can’t be accounted for without making the framework unusably complex, then “it depends” is just intuition that needs to be developed over time and with experience, and is too complex to be codified in a framework (which could end up being the case).
If it CAN be codified, though, doing so would be very useful since it would make the lives of less experienced designers a lot easier, make the dashboards that they design better, and make conversations among designers less confusing (and, I suspect, less controversial). Essentially, I think that it’s worth at least attempting, even if there’s no guarantee that a codified framework could adequately capture “it depends”.
Could also be completely off-base, though. Very interested to know of any holes that you see in this line of reasoning.
Nick,
I certainly don’t think you are completely off-base, and maybe establishing clear guidelines (if not a taxonomy) will help guide novice designers in helping them find and relate important findings in the data.
I also didn’t mean to say that a dashboard *should* be boring; I think it *can* be “boring” and that’s fine; it’s what you do with findings next that takes us into storytelling, presenting, evangelizing, etc.
BTW, you *completely* won me over in your talk when you described a lot of your work as “boring… but with highlights.” I think those are some of the very best dashboards as you make it easy to find the things that matter.
Steve
Thanks for the comments re my “boring with highlights” approach (high praise indeed coming from you) and, oops, I should be more careful in my paraphrasing. *Can* is, indeed, not the same as *should*. I suspect that the underlying point still remains, though. If we say that some dashboards can be boring and others can be more visually engaging, it would be valuable to novices to follow that up with a simple way to determine *which* dashboards should be boring, and which ones should be visually engaging. As I’m sure you do, I see designers making that decision in… well… “sub-optimal” ways all the time, i.e., they try to “spice up” a dashboard that, based on its purpose, would be more effective with a “boring with highlights” design, or vice-versa. Such problems would be less common if there were a relatively simple way to figure out when the various best practices that are floating around apply and when they don’t –if it’s possible to make guidelines for such decisions simple and fairly robust which, I agree, is still an open question.
Nick,
Here’s the path I went down and know that many others have as well.
1) I’m working hard to make these dashboards, and I think they are great.
2) People are not using my dashboards. What do I need to do so they will use my dashboards?
3) I know! I’ll really spice ’em up!
4) OK… that didn’t work (although it did get their attention for two minutes)
5) What now?
File all of this under your wonderful opening slide
Most. Dashboards. Fail.
In my case I have found that the trick to engagement is personalization; that is, make the dashboard about the user as much as possible and show her/him “here’s where you are… here’s where your peers are… here’s where you want to be.”
It also changed behavior (in a good way).
Steve
Thank you for this article Steve. Some of us work in a reporting environment. And all we do is create reports, not discovering anything, not making stories, not really analyzing the data, just simply reporting. With tools like Tableau it becomes a lot easier for a few of us, reporters, to start stories. Thank you for your explanation and pointing out this example: “in a world of 100, 1 will be staving”.
That last word is “starving” not “staving”.
Great article, though I’m a bit confused about something… SB42 was in 2008, not 2007. Am I missing something?
Russell,
The Super Bowl was in early 2008 but it was the 2007 season.
Great insights! Along these lines, I’ve found this talk to be very inspirational on how and what should be communicated: https://www.youtube.com/watch?v=WP-FkUaOcOM
Hey Steve,
Thanks for surfacing the oyster shucking quote of Cole’s. I’ve been having this conversation at least twice a week without the benefit of the metaphor.
[…] Bartlett Iron Quest – Quantified Self Neil Richards How do you visualise an album? Steve Wexler Should you use dashboards to tell stories or to find stories? Kevin Flerlage Avoid the Current Month Drop-Off Lindsay Betzendahl Details Matter: In healthcare, […]