The Art of Data Dashboarding
There is just as much art as science that goes into creating a data dashboard. A well-constructed dashboard always sits in the centre of a triangle of complementary disciplines, uniting data science, graphic design and business consulting. What we call the ‘art’ of dashboard creation sits at the intersection of these three disciplines, and is what this article and its sequel (next month) will uncover, by speaking with three of Dapresy’s most experienced dashboard specialists: Sebastian Öhgren, Client Operations Manager, Fredrik Österberg, Chief Revenue Officer, and Alexander Skorka, Chief Operations Officer & Managing Director, Dapresy Germany.
Part 1: Defining and Designing the Data Dashboard
In part one, we look at the process of defining the data dashboard, deciding what should go into it, and some of the techniques used to bring it to life.
According to Fredrik Österberg, identifying the audience and the purpose for which they want to use the dashboard is one of the most important parts of the dashboard design process. He explains: “For example, in any organisation, one large user group is often management and they will use the dashboard to get a quick overview.” Fredrik continues: “Another group could be the sales team. In this case they may need something more exploratory, it may not need to be as visually appealing, but it could provide some competitor intelligence from surveys. Then the marketing department will want to know the outcomes of campaigns and the return on investment. So even if the source of data is the same, there will be different people that will use the dashboard in different ways and that is what you have to define first.”
Identifying the audience
Sebastian Öhgren adds, “It may be that at the early stages you don’t necessarily know what [the users’] needs will be, and even if you do, that still might change during the process of developing the dashboard. If there are different user groups using the same data, we need to see if their uses align and maybe we can build something that covers both sets of needs, or it may be that they are completely different,” and that could result in different views being provided, or even different data dashboards.
Another useful starting point for Sebastian is to look carefully at the questions in the survey, and understand what the purpose of the survey was. He continues: “It could often be that different groups of questions are there to serve the needs of different groups of users.” Sebastian has worked with clients who have been very clear about what their goals are, and understood the different groups and their use cases from the beginning. On the other hand, he says, “some of our clients just say to us they would like to visualise this part of a survey, and you then need to work harder to get that understanding. But we have a large library of dashboards already created, and we also have an understanding of what might work, so we can then come up with some suggestions and examples for them to see.”
For Sebastian, it is essential that groups of users should get to see a working prototype early, as this will then help them to understand whether the dashboard is going to provide what they want, or if it needs to be refined further. He favours this kind of iterative development where possible, moving from example to a working prototype, and then seeking suggestions on what to improve.
Again, experiences and timescales in different organisations can vary widely. He reports that some complex multi-audience data dashboards have cycled around this process three of four times over a six month period, while other simpler ones have gone straight from wireframe (an outline design without any styling applied) into production in just a few days.
Teasing out all of the use cases seems to happen best when it is a co-creation activity, and users have the opportunity to make suggestions both before and after a new version is provided to them. Ideas often evolve, reports Sebastian. “Sometimes how it ends up can be very different from the way it was imagined from the start,” he says. “There could be a series of small changes, but sometimes we completely change direction. Drastic changes are rare, but it does happen.”
Common purposes
There are several overall purposes that very often occur with business dashboards. Though many organisations or teams within them can have very specific purposes in accessing the data, here are some of those that are more widely found:
Overview – To provide managers, possibly senior managers, with an instant snapshot of what’s happening in the company.
Comparative – To show current performance against targets, benchmarks or past performance. This can result in a classic KPI data dashboard, or against data points from different sources. Dashboards that show trends can also be considered comparative in their nature, as they are comparing the current set of data with previous periods.
Exploratory – Some groups of users will want to explore data in depth, to answer questions or do research (such as the dashboard of sales prospects Fredrik mentioned earlier).
Explanatory – As an extension to the overview or performance dashboard, it is often not enough to simply know what has happened, but to try to understand why. This can be provided by organizing the data so users can drill down to see more observations around what is happening and what may be influencing or causing any recent changes.
However, these are not discrete categories, and one dashboard may include elements that meet
more than one purpose, or are organised in such a way that different regions of the dashboard move the observer from one purpose (e.g. an overview) to another (e.g. explanatory).
Developing the right visual language
“As a company, we are all about visual design,” says Fredrik. “Definitely, one purpose of using graphics is to make [presenting data] more engaging and fun. We have often seen this when data dashboards get used more because they are visually appealing.” For him, the other purpose is that “you can present the data so it is clearer to the user.”
He gives the example of a classic Business Intelligence (BI) dashboard that may present a lot of charts side by side. “A BI dashboard [Figure 1] can be very informative,” he says, “but to find out the reason why you are below target that month, you have to go through all the charts and really understand how the different metrics affect each other. In our version [Figure 2] of this we use visual indicators to make it very clear to the user what the problem is and then provide a clear visual path to the root cause of the problem. This also makes it easier for persons that are not used to working with data to understand what actions they should take to make an improvement.”
Figure 1: A traditional Business Intelligence Dashboard showing a collection of separate charts
Figure 2: The same information presented in a Dapresy storytelling dashboard using visual elements to lead the viewer through from understanding to taking action
For Sebastian, the starting place in deciding how to visualise the data is working out the priorities for the user. He explains: “If there is a KPI that is more important than the others, then that should draw your attention more. That needs to go at the top, in the middle. If it’s a very busy page then what’s important needs to be visually bigger.” Alexander Skorka says that doing some research into what target users are already familiar with can pay dividends.
“Check if there is any learned visual communication already in place in the organization. Don’t reinvent the wheel if you can use the visual vocabulary the target group has already learned and is familiar with.”
Start out with the familiar
This can apply to choosing the chart: “bar, column and pie charts are still the most powerful
visualizations because they are well known and easy for most people to understand,” Alexander notes. Using an appropriate visual vocabulary can also apply to the choice of colours and shapes. He cautions: “use learned colour schemes, e.g. red for negative, green for positive, and do not make your visuals too colorful. Try to help the user to orient him- or herself by applying colour cautiously.”
He also stresses the need for keeping visualisation simple and purposeful. “Visualization is about comparing things like rank, frequency, structure, correlation or the effect of time. Choose what is the most relevant comparison to show and do not try to visualize several dimensions in one visual. People will struggle with complex visual language, so keep it simple. Choose an appropriate chart type and focus on that preferred comparison.”
Alexander also speaks of the need to establish an information hierarchy and identify the relative importance of different items of data, as they will appear to the end user. He observes that many people in research start with the premise that they need to present everything. He explains: “Instead, try to find the right balance between comprehensiveness and compelling visualisation. Let go of the idea that you must show all the figures, and instead try to focus on the most relevant information. It is always possible to provide more details through drill-down, or even self-service components accessed from the dashboard such as a cross-tabulation tool.”
Sebastian Öhgren considers that the visual language adopted by the dashboard must also reflect the culture of the organisation and sometimes the local culture of the country where the organisation operates.
“The portal needs to look as if it is the organization's own thing. They may not have infographical elements on their corporate website, but you can still use the colours and apply some of the graphics in the background. And we’ve noticed that a more informal, cartoonish style of visual seems to work well in America, whereas something a bit more corporate and sober tends to work better with organisations in Europe. The best thing is to get design guidelines from the client and try to make it similar to their own branding.”
Fredrik Österberg sums up the best design process as being one in which Dapresy’s consultants and designers have been able to move the client on in their understanding of what they are trying to do with the data. He explains: “When someone is building an online data dashboard, we find they almost always start out trying to create a PowerPoint delivery system online, and that is not what you should do! Instead, you should be creating a view that focuses on the most important things for each user, and then let them drill down into the information. I really think that is the essence of how to design a good data dashboard.”
What's Coming Next?
In part two, we will hear what our experts do to ensure that the dashboard gets used, and delivers on its promise to drive change and action across the organisation adopting it. We will also look into how to make data dashboards work just as successfully for ad hoc or one-off projects.
