BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
When I hear the word dashboard, I think of getting pertinent information about my car at a glance. How much gas do I have, how fast am I going, what was my mileage on this trip? And is the engine finally warm enough to crank the heat? Similarly, an effective business intelligence dashboard should tell me key things I need to know about my business, in a flash. At the least, a BI dashboard should get my attention, whether the data being presented is good or bad.
I've seen well-designed dashboards that give business users exactly what they require to track business performance, analyze data and make decisions. But I've also seen lesser ones that make me ask, "What am I looking at right now?" What follows are tips on how to design BI dashboards that present useful information to users without overwhelming them with unneeded data.
As with any BI project, planning is the most important step in designing effective dashboards. But many BI teams jump right in without giving it nearly enough thought. Often they don't even involve the business when mapping out what they will deliver; they simply build what they think is relevant to end users. Business managers and executive sponsors have been major participants in every successful dashboard development project that my company has managed for clients.
Up-front planning interviews with business representatives should uncover the key business metrics they need to track and how often they want information to be updated -- for example, monthly, weekly or daily. Other questions to ask include whether there's internal consensus on the definitions of metrics and whether users need to see data on a companywide basis or by region, business unit, product or some other parameter -- or perhaps a combination. The answers to such questions will help guide the design process.
Info overload makes BI dashboards tough to take
One of the keys to good dashboard design is making sure you don't overpower users with too much information. How much is too much can vary depending on job roles. But for many users, it's enough to present high-level metrics that indicate whether business operations are performing well or not. Our planning interviews with clients have taught us that summarized information is what business executives typically are looking for on a dashboard screen, as long as they have the ability to easily drill down into detailed data on metrics that pique their interest for deeper analysis.
As a result, our dashboard design team follows these guidelines:
Limit a single dashboard page or tab to no more than four to six metrics. Also, put the metrics for each subject area or business unit on a separate page/tab. An executive dashboard typically provides a global view of an organization but with many pages and tabs to make the information easier to digest.
Enable users to drill down into data, but no more than three levels. The amount of data necessary to go beyond that isn't conducive to meeting the performance levels expected by typical business users. If they still have questions after drilling down three levels, corporate executives and business managers can call on data analysts or someone on a BI or IT team to get the answers they need.
Use charts, graphs and simple tables that don't create too much "noise" for users. Good choices are bar charts, bullet lists and sparkline graphs that have a visual view of data. Avoid pie charts; in general, the human eye is good only at comparing slices that are right next to each other. Bar charts are much better for conveying differences in BI data to users. Gauges also are overrated for dashboard use. They're extremely effective in gaining the attention of users and can be the sizzle in selling business intelligence dashboard projects internally, but they present little information and take up valuable screen space.
Make dashboards interactive so users can get more information and customize data views. In addition to drill-down capabilities, dashboards should contain filters that enable users to narrow or expand the scope of the data being presented to reflect their individual information needs. Alerts can also be used to call immediate attention to information based on pre-defined criteria set by users.
In designing dashboards, we don't follow a traditional waterfall development process -- we always use Agile development methodologies. Our usual practice is to start by creating a "straw man" design to give business users an idea of how a dashboard would look. After we gain acceptance for the basic look, we develop a prototype, get feedback from users -- and go back to the drawing board.
Final word on dashboard design goes to the users
Truthfully, there hasn't been one project where we've nailed it the first time at the prototype stage. As well as you plan and design these things up front, nothing should be set in stone until the users see the end product of your work. It's important to understand that this is a process and there are likely to be several iterations before you settle on a final design for your business intelligence dashboard.
For more on dashboard design
Learn how to de-clutter dashboards for effective design
Take a closer look at the evolution of data visualization
Watch a video on effective dashboard and scorecard design
And once a dashboard has been finalized and put into production, the next thing you know, someone is asking for more detailed information or additional data from another source system. BI, reporting and analytics is a constant work in progress. As a result, dashboard designs are, too. You're likely to deliver functionality and data in ongoing phases; the key is to get wins in each phase.
Ultimately, dashboards are the surface of all of a company's BI data. Building and updating them is a long march, but if you stick to effective dashboard design principles and practices, your users should get what they need to make better business decisions.
About the author:
Brian Jordan is vice president of sales at TriCore Solutions, a consulting and managed services provider that focuses on business intelligence, data warehousing and enterprise resource planning projects. Jordan has more than 15 years of BI experience in a variety of consulting, technical sales, operations and management positions.