The Dangers of “Easy” Open Data Dashboards

By Frances Ruiz

When we launched our budget dashboard in May, we had solved many of the technical challenges of creating public dashboards from open data. We can now publish a page of interactive charts and tables in a matter of hours.

But while it’s easy to splash up interactive charts by the dozens, it’s actually really hard to create a successful dashboard. Which charts are best to use? How should they be arranged?

At first we fell into the trap of churning out charts without stopping to think explicitly enough about what we wanted to say. The result: unclear messaging, lots of rewrites, audience confusion, and time-consuming miscommunications between departments, each with their own ideas about the purpose of the dashboard.

Now, before we design a single chart, we answer the following questions:

Who is the audience?

There are many potential audiences for each dashboard. For our homelessness dashboard we identified the following primary groups:

Each audience brings different background knowledge and contextual understanding of the data. It is impossible to create dashboards that will give every audience exactly what they are looking for. Our open data dashboard efforts aim to deliver performance metrics and value to the general public. We additionally include separate data pages where data driven investigators can explore the data more deeply and gain even more value.

What are we trying to communicate?

Generic reasons might spring to mind: “We want to show we achieved our goals”, or “We want to show how taxpayer money is being spent”.

These broader goals are a good start, but we have to drill down to specifics. What should people learn from going to this dashboard? Which concrete questions are we answering? Some questions for our homelessness dashboard might have included:

  1. How many children are homeless in Asheville?
  2. How successful have local programs been at housing veterans within certain timelines?
  3. Why do the numbers of homeless people in the Point-in-time counts never go down?
  4. With these questions in hand, we can proceed to designing a dashboard around them.

Are we providing enough context to understand the data?

Data visualizations are meaningless without context. In our homelessness dashboard, we tried to address certain contextual misconceptions that service providers often encounter. Service providers continually hear, “The point-in-time counts are always the same each year, so that means no one is doing anything about homelessness!” In reality, there is a lot of turnover in the homeless population – the counts do not reflect the same set of individuals every year. To provide this context, our dashboard explains that there is a large influx of new homeless people to the region every month.

Monthly inflow and outflow of homeless veterans

Are we giving our audience what they need?

I can theorize about what the “average” citizen is curious about until the cows come home, but unless I actually ask people from the intended audience, I’m likely to get it wrong. At the City of Asheville, we are making a concerted effort to start user testing with a diverse pool of users early in development and to continue engagement throughout the process. In fact, in the earliest stages we can let users help us figure out what the questions should be simply by asking: “What do you want to know?”.

As we emphasized in our last post, dashboards are a process, not an endpoint. As we build out these early examples, we are learning better ways to develop them. In a future post, we’ll talk more about what happens with them after launch.

Originally published August 10, 2017