Dashboards: Another Look at Data Visualization

written by Michael MacDonald, CEO, Chairman and Founder, Visual Mining, Inc.

Executive Summary
Data visualization applications have traditionally shown poor ROI but are finally starting to show real value in dashboard applications. The key is closely integrating business goals and processes with the right technology.

Background

At a recent meeting in Boston, a pair of industry analysts told us that they were reluctant to publish an update on data visualization applications. They had talked to a number of software vendors, but couldn't convincingly articulate how those products could help a business's bottom line. Years of expensive, over-hyped products with underwhelming deployment results had jaded them and their subscribers.

We believe their perspective stems from a common misunderstanding of visual technology's role and capabilities. We've been involved with a number of data visualization projects that have produced positive results without breaking the bank. Data visualization applications can save money and produce organizational efficiency – it can work.

Making data visualization applications work successfully comes from mapping real business goals and processes to the right technology.

An example of a dashboard that tracks progress
of fundraising campaigns for a non-profit.

A Dubious History

Most people don't need to be convinced that data visualization is useful. All kinds of studies have shown that we humans respond well to visual data. Graphics catch our eye and, if the picture is familiar and well designed, we can digest the information quickly. Manufacturers effectively use graphics to prevent little problems from becoming big problems; financial types spot trends or opportunities that earn or save them millions of dollars; medical services can remotely monitor patients, preserving good health and saving lives.

So why do analysts and CIO's remain skeptical? Because the process of creating relevant graphics from arbitrary data sources, to integrate with ever-changing deployment environments, is not trivial. Most software packages are unable to address client data visualization needs out-of-the-box without at least some custom work being done by the client's IT staff or the vendor's professional services group. Too often that basic fact is obscured in the face of slick demonstrations with cool graphics and feature-laden development tools. Not accounting for the true costs involved and benefits received has led to many multi-year, budget-busting, failed projects.

Another Look

But even once-burned skeptics should take another look at graphical applications, and at dashboards in particular. The technology has improved, and the process is better understood. There are a number of successful dashboard projects that have been able to deliver real value for a fair price. The key is in knowing how to implement them.

One of our favorites is a performance metric dashboard in use by the US Air Force's health services executives. The Air Force runs approximately 80 clinics and hospitals worldwide, organized into a multi-tiered command structure. Performance metrics, such as vaccination percentages and cancer-testing rates, were established to track adherence to headquarters-directed goals. A series of interactive pages were designed, incrementally, to compare each health center's current and historical performance against the target, and against other clinics. Then, in a monthly videoconference, senior Air Force executives commend those that have embraced organizational priorities and demand accountability from poor performers. The graphical presentation of the information makes it plainly obvious to all where each clinic and organization stands. This application has been credited with saving both taxpayer’s money and service men and women's lives.

The Air Force Medical Services use performance dashboards
like this one to monitor and hold accountable more than 80 clinics
on their adherence to mandated performance targets.
See the AFMS Dashboard here. (click on the "Go" button at top right to see a dashboard)

Lessons to Take Forward

There are a number of lessons to be learned from the Air Force example. The most important is recognizing that understanding the goals and processes is far more relevant than the selection of the technology. No software could drop into that project until the goals and measurements toward those goals were well defined. Another critical factor was to incrementally build the application, which allowed the data flow to be tested and end-users to view and adjust the application using real data in a real-world environment before the entire project was implemented. Finally, the monthly follow-up was, and continues to be, a critical component in the project's success.

Understanding the Technology

The careful selection of data visualization technology can save your organization hundreds of thousands of dollars and months of development time. Minimally, the technology must be able to extract data from any source, generate the desired graphs and tables from that data source, and then be able to integrate those graphs and tables in context with other application information. Better packages will also enable users to easily navigate through data to detail information or other documents.

There are hundreds of overlapping data visualization tools available, from simple charting components, to graphical reporting suites, to full-blown business intelligence tools. Which one you select should be based on your resources and your needs.

Most of the vendors will allow you to evaluate their product before buying, and you should. Forget poring over requirements documents and feature lists. Forget the RFP's. Don't assume that simply because the vendor is large or because the product has tons of features that it will work in your environment. See if you can build real examples using real data before buying. See if the end-users readily understand the visualizations. See if you can put the visualizations in context with other information. See how much vendor involvement is required. If you can't get some basic, relevant graphics up and viewable by end-users within a couple of weeks, you may want to consider another vendor.

Best Practices

Below are a number of suggestions gleaned from hundreds of successful implementations that will help you succeed in your next data visualization project.

Focus on the goals. This is a simple, straightforward rule that has gotten lost or marginalized in every failed visualization project. Don't be distracted by the graphics or the technology - focus on the results needed by your organization and end-users.

Know your data. Most non-technical people don’t understand that the hardest part of "data visualization" is preparing the data, and not generating or deploying the visualization. Graphics and other data visualizations often require the access, coordination, integration and aggregation of multiple data files, databases or other sources. No data visualization project should be started without knowing who will do this step and how long it will take.

Understand the process. Producing a good data visualization usually involves several steps, each of which must be understood. How does data get into the system? How does the data get compiled into usable form – and how often is it updated? When do the graphics get generated? Who will use the system? Who verifies the results? How do users actually use the system?

Don't buy features you don't need. If both the $100,000 system and the $10,000 provide you the same features that you need, then why spend $100,000 purchasing software for features that few, if any, users will actually use? Make sure you tie the software's features back to your goals. Features that do not contribute to the goal are often costly and unneeded. Aim for product flexibility, not features.

Make relevant, understandable graphics. What question does the graph answer? The graphic should allow the viewer to quickly see what a table of data cannot show at a glance. Simply reinforcing textual data with a graph, or using a graph where a few words would work equally well, is a waste of time and money. Stay away from 3-D effects and animations unless they contribute directly to the meaning of the graph.

Keep the graphics in context. Simply showing a graph alone on a page rarely addresses the questions that it should answer. How does the user get related information? Should the graphic show both historical and recent information?

Take an incremental approach. Too many projects fail under their own weight of requirements documents and general balkanized institutional paralysis. Data visualization projects often affect different organizations differently, so the sooner you test your infrastructure, the better you'll be able to identify potential problems. Providing real results early on in the process, as well as continuing improvements, can work wonders to gain buy-in from your end-users.

Follow up. Ensure that people are using the tools available to them. The selected technology should be flexible enough to tune to changing requirements and conditions.

Summary

Data visualization can be a tremendous asset to nearly every organization – however, without clear knowledge of where you’re going, the technology won’t be able to take you there.

Download the PDF of this article.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

About the Author

Michael MacDonald, CEO, Chairman and Founder of Visual Mining, Inc., has extensive experience in software design, development, and implementation, specializing in information graphics and Graphical User Interface applications. Recognizing the need for easy-to-implement dynamic graphics as well as the corporate support for the Java standard from Netscape and Sun Microsystems, he correctly assessed the value of a dynamic web-based graphics solution, and the need to supply chart components for developers and HTML authors.

Before starting Visual Mining, Mr. MacDonald's work was incorporated in several large-scale software applications from AT&T, Intersolv, Hewlett-Packard, Tracor, and Shiva Corporation, as well as numerous applications within the US government and military.

In 1995, Mr. MacDonald started MJS Consulting, a computer services organization providing software support to government and Fortune 500 companies, including Sun Microsystems, Fannie Mae, National Security Agency, Central Intelligence Agency and several others. He has also worked for Rapid Systems Solutions Inc, Tracor Aerospace ISD and the National Security Agency as a senior research intelligence analyst.