source: Information Management Special Reports, February 24, 2009
Business intelligence is a rapidly growing sector of IT development and expenditure budgets. Products include financial modeling, data analysis, data mining, reporting and charting tools. The large vendors, Oracle, IBM and Microsoft, all have products in this space and continue to develop new ones and refine the existing ones. These tools generate huge amounts of additional data and charts about the business data. This, in turn, creates a new problem: how do you make all this data readily consumable and, more importantly, actionable? How do you summarize the data? The answer is a BI dashboard.
Most of us look at a dashboard in our everyday lives, in the car. When looked at closely, the car dashboard provides a perfect model for a BI dashboard. Some information is updated in real time, for example, traveling speed and the engine speed (RPMs). Some information provides warnings, such as low fluids or worn brake pads. Some reflects the current state of something, like drive or park. Taken as a whole, the dashboard provides a quick and easily consumed overview of the state of a car. Some cars provide an additional level of detail, for example, a mileage range for the car based on the amount of fuel remaining. Behind this is a real-time display of MPG and an average MPG, which allows the user to drill down to gain more insight into the original piece of information, the range. By drilling down and looking at the data behind a number, a user can gain additional insight and answers to questions. A BI dashboard acts the same way: it provides a quick and easy way for the user to see the state of “the business” and may provide the ability to drill into the information to provide detail. To be successful, the dashboard must present the information in a straightforward fashion in the form of charts, traffic signals, gauges, dials and, in some situations, spreadsheets.
The BI dashboard converts numerical information to charts and gauges, or color-coded graphics, which can be efficiently perused and understood. For example, a business may have a key metric that needs to be within a range: between 1 to 10 is good, 11 to 25 is ok, and above 25 is bad. If the metrics displayed as a numerical, the range info must also be provided so the user can put the value in context. This takes up screen real estate and takes time to for the user to process the information. In contrast, a key performance indicator (KPI) in a BI dashboard for the same metric, a value of 1 to 10 is graphically displayed as a green graphic, 11 to 25 as yellow, and above 25 as red. The user does not need to know the range values; green means that value is where the business has defined it needs to be.
Allowing a user to process information more efficiently is only one advantage of the BI dashboard. The benefits of these tools can have far-reaching, enterprise-wide effects. There are several reasons for investing in the design and construction of a BI dashboard. It maximizes the ROI on the BI analysis tools, it reduces decision-making time and increases business agility, helps the entire organization to focus on what the key metrics are for the business and, finally, can increase quality of information used to make decisions.
Speeding and Increasing the Efficiency of the Enterprise
In the past, the available data analysis tools required human effort to convert the analysis to charts and graphs. Once formatted, the analysis could be disseminated for decision-maker consumption. This time-consuming process restricted information flow, as it was generated quarterly, monthly or perhaps bi-weekly. Business performance measurements were therefore restricted to the same time frames. Consequently, not only were decisions made based on old data, but the reaction time of the business was measured in weeks, at best. By converting the human effort into “coded process” in the BI dashboard platform, the results of the analysis can be made available on demand, with the data being as up to date as the data sampling frequency - allowing the decision-maker to benefit from accurate on-demand data analysis.
Focus and Data Quality
A BI dashboard focuses its users on the key metrics that drive the organization forward to success. An organization can monitor the key metrics and react quickly to any changes, reducing the latency between an event and the appropriate response. The dashboard can then become the central point by which the organization or team can monitor its performance. Its creation can fundamentally change the nature of the information used to create it. If reports were created monthly, then the data was likely collected monthly. By allowing the data to be displayed on demand, users will naturally desire the data to be refreshed more frequently. This will soon reach equilibrium when the collection or sample rate reaches the optimum rate for the data source. By increasing the data-sampling rate, the result of the analysis can be improved by providing more data points for the analysis, for example, when averaging the temperature over the span of a month, versus sampling just one day.
Dashboard Design Variables
Dashboards are designed based on the requirements of the end users. The two basic forms of dashboards are business reporting and operational. End-user needs then determine how frequently the information is updated. The two modes of operation available in a BI dashboard are snapshot and real time.
Business reporting dashboards are typically used to show how the business is performing over a specific period of time. Reporting dashboard usually contain financial information, such as revenues, forecasts, projections and tracking information. They may break down sales by product or region. In some larger organizations, project rollup information is displayed for major strategic initiatives, providing senior management an enterprise-wide view.
For example, financial services companies build dashboards to display summary market information for global markets, displaying such things as GDP growth, population trends, employment demographics and economic data. These dashboards are maintained by analysts and are used by traders and other analysts within the organization. These kinds of dashboards tend to operate in a “snapshot” mode because the data does not change that frequently.
E-commerce companies frequently use real-time business reporting dashboards to provide hourly or daily feedback on sales. Such dashboards enable new marketing copy and offers to be posted, and feedback to be quickly gathered and reported. This speed of reporting allows companies to be extremely agile, constantly trying new approaches. The use of real-time reporting is not just limited to e-commerce business; some manufacturing enterprises use them to report on products manufactured and shipped.
Many scenarios in business can benefit from a real-time operational dashboard. Real-time dashboards are sometimes not built on top of a traditional BI platform; they can be built using standard Web development technologies. Operational dashboards allow the organization to manage intraday events and operate at peak performance by identifying changes in the operating environment. For instance, an operational dashboard for an Internet-related company would allow the company to monitor its server farm and load. The dashboard would display items such as the number of servers running, the request queues per server, network bandwidth use, CPU load, number of active user sessions and average request duration. This allows IT staff to quickly scan the dashboard and know that the site is operating within defined limits and that nothing unusual is happening.
Designing the Dashboard
A good practice when starting a dashboard project is to first build a prototype that uses easily obtained data or demo data. This prototype should be used to promote awareness and the benefits of a having a dashboard.
Once the decision has been made to construct one or more dashboards, the first step is to identify the key metrics that are to be displayed and the manner in which they are to be rendered; this can be a time-consuming process. Some of the information will be displayed as traffic lights, gauges, charts or graphs. Some may also be spreadsheet data.
Having identified the metrics to be displayed, the next step is to identify from where the data can be obtained. This can be challenging because the data may not currently be gathered or readily accessible. In general, it is better to release a dashboard that is missing some metrics rather than wait for the challenging metrics. Releasing it early allows business value to be realized as soon as possible.
The next step is to build the data gathering and processing layer. In some cases, the results from the BI infrastructure can be used. Other metrics will require new code development. By designing with changes in mind, companies allow for greater business agility, because the dashboard can be modified quickly to meet new business requirements.
The final step is the actual dashboard construction. The dashboard should ideally be constructed as just a presentation layer, with minimal business logic.
Constructing the Dashboard
The type and mode of a dashboard will often determine the tool that is used to build it. As discussed earlier, operational dashboards can often be built using standard development technologies such as Java or Microsoft .NET. These tools allow developers to sample data from a wide variety of sources, such as performance counters generated by the underlying operating system and external monitoring devices. The data sources can then be combined with rich graphical display tools for charting and gauges to build both snapshot and real-time dashboards.
Business reporting dashboards are typically built using the dashboard tools from the same vendor as the BI tools, although a mix-and-match approach can be used. Some products use proprietary development languages and tools to build the dashboards. Adopting these tools should be considered carefully, since the long-term cost of ownership could be much higher than adopting tools that use development languages that the enterprise has existing expertise in.
There are three critical success factors for a dashboard: metrics, sampling rate and platform. The dashboard must provide the right metrics, sampled at a business-appropriate level and rendered on an agile platform.
Ian Dicker is director of Architecture at Syrinx Consulting, based in Waltham, MA. He works closely with Fortune 1000 companies in the financial services, pharmaceutical and Internet sectors to optimize collaboration platforms to meet business objectives. Dicker blogs regularly on reporting, collaboration and architecture best practices and speaks at developer conferences. A native of Great Britain, he graduated from the Thames Valley University, Slough.