Data Analytics

IT Managers Can Facilitate Good Decisions with Strong Data Visualization

Building a data dashboard requires careful use of data and statistics, and knowledge of the audience for them.

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.

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

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. State and local IT teams can use visualization tools and dashboards to help both managers and customers pull meaning out of data previously locked away.

But there can be severe consequences to choosing the wrong statistics or the wrong graph style, leading to constituents and managers making 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 some best practices to get started.

1. Know the Data to Avoid Errors in Graphic Representations

Low storage costs and inexpensive, fast, database engines make it easy to save raw data from many different sources. Documenting the metadata that make the raw data useful can prove more time-consuming. Without context, however, a raw data stream is open to misinterpretation.

For example, electric meter usage data can help in planning for energy management and maintenance programs. But someone using that data could easily come to incorrect conclusions without documentation that the data were available only from a sample of homes that have had their meters replaced. Those meters may account only for new homes, specific neighborhoods or young families.

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, there cannot be certainty that the visualization gives a good picture of what is really happening. People may draw false conclusions.

Documentation of data sources and context and creation of good metadata are critical first steps to any data visualization and dashboard project. Providing context for graphics or the dashboard is equally important. Various tactics can put things into context for the viewer — including minima/maxima, comparisons to previous data and overlays of related data.

2. Know the Audience Whether Decision Makers or Constituents

Data visualization tools from vendors such as Microsoft, Oracle and IBM make it easy to 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 — it requires knowing the audience that will use the visuals. Do they use this data to make decisions or do they simply monitor it? Do they 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 constituents to see on the web?

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 whether to create a problem report 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 together on the same visualization.

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

3. Know the Statistics to Present Accurate Information

When a dashboard simply shows time-based data on a graph, it doesn’t involve a lot of statistics. But when a visualization tries to show the relationship between different data sets through correlation, summary and averaging, or other statistical manipulations, it’s important to ensure the math is right. 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 their 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, treemaps 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 governments to help end users understand data is to use geographic-based visualizations, since so much of what governments deliver is based on location. When using geographic visualizations, however, it is important to use clear, precise geocoding. For example, if data were coded only to the county level, then users shouldn’t be able to zoom into a level that implies greater precision.