Data Analytics

3 Best Practices to Consider When Building a Data Dashboard

Visualization is more than just a pretty picture; your audience must understand the context of the information as well.

Joel Snyder, Ph.D., is a senior IT consultant with 30 years of practice. An internationally recognized expert in the areas of security, messaging and networks, Dr. Snyder is a popular speaker and author and is known for his unbiased and comprehensive tests of security and networking products. His clients include major organizations on six continents.

In an environment where politics and special interests often control the agenda, good data-based decision-making offers an objective way to allocate resources, respond to complaints and comments, and deliver services to constituents.

Federal IT teams can use visualization tools and dashboards to help both managers and end users pull meaning out of data previously locked away.

However, there can be severe consequences to choosing the wrong statistics or the wrong graph style, leading to bad decisions. IT teams shouldn’t be ashamed to ask for help and guidance to ensure that data visualizations are accurate and useful. Here are three best practices to get started.

1. Understand the Raw Data the Agency Takes In

Data visualization often brings together different data sources in the same image to show correlations. Without a good understanding of the raw data and its meaning, a viewer cannot be certain that the visualization gives a good picture of what is really happening.

Without context, a raw data stream is open to misinterpretation and false conclusions. For example, COVID-19 vaccination information is being eagerly watched right now. Someone using that data without documentation could easily come to incorrect conclusions about whether the data covered individual injections (even if they were for one person) or was counting individuals needing two injections as one.

Trouble could also arise if the vaccine data only included vaccinations administered by state programs and didn’t include multistate programs, such as the one sponsored by the Department of Veterans Affairs.

Knowing your data is the first step to avoiding errors in graphs. Low storage costs and inexpensive, fast database engines make it easy to save raw data from many different sources. Documenting the metadata that makes the raw data useful can prove more time-consuming.

Still, that documentation is a critical first step to any data visualization and dashboard project. Providing context for graphics or the dashboard is equally important. Various tactics can put information into context for the viewer, including finding the minima/maxima, creating comparisons to previous data and overlaying of related data.

2. Identify How Your Users Leverage Data

Next, make sure you know your audience, whether it’s agency managers and other decision-makers or end users. You have lots of options when using data visualization tools from vendors such as Microsoft, Oracle and IBM that can help pull together data from disparate databases and quickly build a wide variety of graphics to present the information.

Picking the right information and presenting it in the right way requires more than knowing the data, however. It requires knowing the audience that will use the visuals.

Does this audience use this data to make decisions, or simply to monitor? Do users dynamically adjust the visualization, or would a static presentation be more appropriate? Does the audience make short-term financial decisions or decide long-term strategy? Are the visuals for internal consumption, or will they be available for agency users to see on the public internet?

2009

The year the first federal dashboard was created

Source: gcn.com, “Recovery.gov: Point, click, follow the money,” Jan. 14, 2011

Making choices on graph types, data overlays and even color depends on knowing what type of decision-making will come out of the data visualization.

When a dashboard supports short-term operational decision-making — such as determining whether to escalate a potential problem or to approve a request — data visualization can speed decision-making and improve accuracy through the use of colors, graphics such as arrows and limit lines, and graphical and table data on the same visualization.

Although it’s tempting to be transparent and deliver the same information in the same format to agency end users as to the public, doing so can lead to significant misunderstandings. The general public won’t have the same context and subject-matter expertise as internal users, and will require a presentation in a different format with significant additional explanations.

3. Turn to Data Scientists and Use Accurate Stats

For a simple dashboard with time-based data on a graph, there are no statistics involved. Anytime a visualization tries to show the relationship between different data sets, however, you’re in the world of statistics — and that’s not always a good place for amateurs.

It’s important to ensure that the math is right when you are using correlation, summaries, averaging and other similar statistical manipulations. A full-time data scientist can help ensure the accurate use of statistics. If a full-time data scientist is not available, some short-term help can ensure appropriate use.

A similar concern lies in the graphics themselves. Many business intelligence and visualization tools have a huge variety of graphic options — not just tables and charts but also bubble charts, heat maps, mosaic graphs, tree maps and more.

Selecting the right graph type and the right axis scales is not just a matter of aesthetics; it’s the best way to focus the viewer’s attention and deliver the right information as accurately as possible.

For IT teams that have learned to graph using Excel, some training on choosing and presenting information in graphics is a requirement. A good place to start is with Edward Tufte’s classic book The Visual Display of Quantitative Information.

A natural way for agencies to help end users understand data is to use geographic-based visualizations, since so much of agency work is based on location. When using geographic visualizations, however, it is important to use clear, precise geocoding. For example, if data was coded only to the state level, then users shouldn’t be able to zoom in to a more specific level that implies greater precision.

They say a picture is worth a thousand words, but the right picture can be worth so much more. It can be deceptive and confusing — or it can explain, persuade, clarify and convince.

GO DEEPER: How can chief data officers help agencies?

FreshSplash/Getty Images