Sergey Nivens - Fotolia
Today's ecosystem of data visualization tools offers users a long list of functionality that gives them plenty...
of control over how to present business intelligence and analytics data inside their organizations. But in data dashboard design, too much of a good thing can often be bad.
At the Tableau Conference 2015 in Las Vegas last week, users of Tableau Software's BI and data visualization tools talked about their best dashboard designs -- ones that avoid information overload and contextualize data to ensure that business executives and other downstream users are able to get useful info from the dashboards.
The first thing to keep in mind, said Meghan Carton, former coordinator of the school funding team for the Washington, D.C., public schools, is to make sure that dashboards are targeted and focused. Carton, who now works for an organization called Polaris that fights against human trafficking, said users of data visualizations often don't know exactly what they're looking for in a data set, so they default to the position of asking for all the available data. But trying to cram too much into a dashboard makes it hard to read, she cautioned. To steer clear of that, Carton said, developers need to work to understand the questions users are likely to want to ask and build dashboards that tackle those specific questions.
Keep it clean on dashboard design
During her time with the D.C. public schools, Carton developed a BI dashboard that lets parents track how the city's education budget is distributed to all the individual schools and how those schools allocate their funding. She and her team started with a single dashboard that was designed to be visually appealing. It had colorful heat maps, tree maps and bar charts, and Carton said it looked great.
The problem, she added, was that it was hard to digest the information. "We said there's no way a parent sitting in a community meeting is going to pull this up and get anything out of it more than they would with a spreadsheet."
So going back to the drawing board, the funding team put together something simpler. One of the best dashboard designs they came up with after roughly 20 iterations has bar charts that break down the specific areas in which schools spend their money. Hovering over a bar pops up more detailed information. There's also a text box that verbally contextualizes the data, and users can click on chart elements to drill down even deeper. And instead of all the data being on one page, it's split up and organized to reflect how people are likely to interact with it. Carton said this kind of simplified, clean approach to visualization and data dashboard design has helped parents go into community meetings to discuss school budgets armed with better information.
User experience trumps design impulses
The other downside of pushing too much data into a single dashboard, aside from visually overloading viewers, is poor performance. During a presentation at the Tableau conference, Jeff Bloomfield, a senior BI analyst at Chicago-based Cars.com, said his team faced that issue in developing a dashboard for the dealer franchises that the provider of car-buying information works with. The analysts wanted to create a dashboard that would allow dealers to track marketing campaign performance and view details about potential customers in their local area. The BI team's first impulse was to converge all of that data into a single dashboard.
But that didn't turn out to be one of their best dashboard designs. Aside from it being hard for users to make sense of the data, Bloomfield said load times were terrible. In an age when people are used to near-instantaneous loading of webpages, expecting users to sit around for 20 seconds or more while the dashboard loaded was unrealistic. So the analysts split up the initial single dashboard into 16 separate views. That alone improved page load times by 40%, he said.
Jeff Bloomfieldsenior BI analyst at Cars.com
Another thing they did to improve load times is to render static graphs as image files, which load much quicker than typical Tableau dashboard components. Bloomfield said taking those relatively simple steps to improve the dashboard's performance has increased the number of dealers who actively use it -- and ultimately, that's the most important thing.
"It doesn't matter how cool the visualizations we build are," he said. "It won't get adopted properly if it doesn't work."
A little color helps make data clear
At travel website operator TripAdvisor Inc., results from daily A/B tests of site layout and design changes are sent each morning to product managers who oversee the tests. But they used to be sent as spreadsheets embedded in HTML email messages, and Eric York, analytics manager at the Newton, Mass., company, said it was hard to look at individual cells of the spreadsheets and know which tests were performing well and which weren't because everything was rendered as simple text.
The analytics team recently switched the daily emails over to embedded dashboards that now color-code successful tests in green, failed ones in red and inconclusive outcomes in gray. Product managers can also see the confidence level in the tests to know which improvements, such as time on page or the number of accommodations booked through the site, are deemed to be statistically significant. York said this may seem like a relatively simple change, but it has had a huge impact on the ability of the product managers to communicate results from tests to TripAdvisor executives.
"There [are] no more questions from the CEO about how we're doing," he said. "We see all of that every day at five in the morning."
Check out more examples of best dashboard designs
Good dashboards should be more about utility, less about visual glitz
Avoid these common mistakes when designing data dashboards