Business intelligence has a terrific ring to it, but you may be wondering what you can really do with BI to deliver concrete benefits to your company. Here, I help you understand how you can exploit BI simply and without a horde of paid consultants to set everything up.

Let's first understand what BI is. BI is a broad term that encompasses managing data to produce "actionable" information. The main steps in BI include planning, gathering, organizing, analyzing, and presenting data. Classic BI produced periodic financial, sales, inventory, and other types of reports. These reports were, and still can be, useful for basic financial information and performance tracking and high-level trend analysis (e.g., company revenue). Standard reports may also be necessary to satisfy regulatory requirements or allow investor overview.

Most information that such reports produce, however, may be useless for many of the decisions that managers and operational staff must make. Often standard reports lack the right information, or the information isn't timely enough. Additionally, the information may be too coarse-grained or cluttered with too much irrelevant detail. And although standard reports typically include basic summary information, such as totals, averages, minimums, and maximums, they may not provide the more sophisticated analysis necessary to establish causal relationships or to make predictive forecasts.

The trend now is for what I call "agile" BI, which focuses on providing the right information at the right time to make decisions. This information is tied to clearly identified goals, is based on sufficiently current and accurate data, and contains minimal extraneous information. Agile BI offers decision makers advanced analysis, including causal relationships and information about leading indicators, which they can use for predictive purposes. (A leading indicator is a measure that has a significant correlation with future performance.) Agile BI delivers information in a variety of forms and is rapidly adaptable to changing needs or evolving source data.

Decisions and Intelligence

To help you understand how businesses can use BI, it helps to categorize decisions as operational, tactical, or strategic. Operational decisions involve production and delivery of the enterprise's products and services — for example, detecting variance in product quality and identifying the cause of the variance. Tactical decisions look ahead to a fairly short time horizon to optimize various goals, such as net income. Deciding which product and customer sets to focus on for the next quarter would involve tactical BI. At the highest level, strategic decisions have medium- to long-range scope and can launch new product lines or service areas, select distribution channels, or drive mergers and acquisitions.

I cover some examples in each of these areas shortly, but first let me explain another dimension of BI — what kinds of "intelligence" it can provide. Customer or client characteristics and their purchasing or service-use patterns lie at the heart of many classic and agile BI systems. This type of information is essential for establishing causal and predictive relationships that enable decision makers to assess alternatives for many key decisions. Similarly, most BI systems encompass manufacturing capacity, production volume, and some measures of quality (or the analogous data for providing services). With a focus on providing the right information at the right time, agile BI broadens the field and often incorporates information about competitor and business partner products and performance, economic environment and trends, and custom survey and research data.

Similarly, agile BI produces information that businesses can use in more powerful ways than they can use classic BI reports. One way to envision this idea is with the following scale of how information is used, ranging from simple to complex:

• reading — learn basic facts
• monitoring — check ongoing status
• trend analysis — analyze time series; observe growth and change
• root cause analysis — drill down to examine source of status, trends, and so forth
• statistical analysis — discover correlative and causal relationships
• predictive analysis — model decision outcomes, analyze risk, produce "what if" scenarios

Classic BI generally supports the first three ways that a decision maker can use information, whereas agile BI covers the whole range.

BI in Practice

Product-quality monitoring has been common practice in the manufacturing sector for many years. Now businesses are applying similar principles to other parts of the business process. This approach is known as Business Activity Monitoring (BAM), and its three key elements are close-to-realtime monitoring, graphical dashboards presenting summary information, and interactive tools to find the root cause for what the dashboards display. For example, one company put a BAM system in place when it launched a regional sales promotion. Shortly after the promotion began, a dashboard view of sales results showed that the peak response to the promotion was occurring between 7:00 p.m. and 9:00 p.m. The BI system also maintained a constant analysis of variance in performance among stores and raised an alert when one store's performance fell significantly below the mean. The manager responsible for the promotion used the system's drill-down capability to look for the root cause and quickly found that the lagging store was closing at 6:00 p.m., rather than the later closing time of most stores. With that information, the appropriate action was obvious: The store would promote longer hours at least for the duration of the sales promotion. In hindsight, this example may look obvious. But with conventional reporting, the "obvious" would probably have been discovered too late or never at all.

The concept of data mining has been around awhile, but only in the last few years has the technique been connected with realtime operations. "Data mining" is an odd term — after all, when people dig in the ground for diamonds, we don't call it "dirt mining." A better term would be "information mining" because the process is basically the analysis of multiple variables to discover useful relationships buried in a mountain of data. One operational use of data mining is to guide cross-selling, both knowing which products to suggest and when to stop so that a customer doesn't get turned off. The sales guidance capability is implemented by embedding a rules engine in the interactive (or web) sales application and frequently updating the specific rules by running a multivariate analysis of the sales data.

Data mining is a good tool for tactical and strategic decisions as well. As an example, a publishing company uses data mining to discover how to tailor subscription-renewal special offers to different subscribers. A step beyond data mining is predictive analysis, in which decision makers use data analysis to build a model through which they can run a variety of "what if" cases to compare expected outcomes. Predicative analysis provides a powerful decision tool for planning direct marketing, customer affinity programs, risk management, and a variety of other tactical and strategic undertakings.

One company used a model to create a tactical plan for capturing competitors' customers. The model was based on the company's sales and customer satisfaction data, as well as data from surveys of competitors' customer satisfaction. The company examined outcomes from several scenarios involving different product lines and promotions in order to pick the combinations that were expected to win over and hold on to the most customers.

Modeling with predictive analytics can help with strategic decisions too. When a major storage and distribution company prepared to set out a strategy for selecting transportation services (e.g., rail, trucking) and business partners, it created a complex model involving data on its own and other companies' revenue, costs, service levels, customer forecasts, and many other factors. Running the model with different strategies and different future scenarios let the company create a strategy with high anticipated return and minimal risk.

Current Trends

As more companies move from classic to agile BI, several trends have emerged. Operational data stores are a common foundation for many BI systems. Basically, an operational data store provides close-to-realtime propagation of operational data to a structured copy that analytic tools can access. The goal is to strike a balance between having true realtime data and having data that's easy to manipulate and has less performance impact on production applications.

Embedded business intelligence incorporates predictive analysis into interactive and B2B applications, as we saw in the cross-selling example. Embedded BI has significant potential for improving both revenue and service for many enterprises.

Balanced scorecards are another growing trend. A scorecard is just some summary measure, such as sales revenue. A balanced scorecard includes measures from the financial, customer, business process, and learning and growth perspectives to provide a holistic perspective on how the enterprise is doing. Balanced scorecards emphasize leading, rather than lagging, indicators. As an example, if an enterprise depends on a highly trained workforce, two "learning and growth" leading indicators are success in hiring and employee training. All other factors being equal, when both of these are low, the enterprise is likely to perform worse in the future.

Balanced scorecards also incorporate causal relationships between goals and indicators and among different organization areas. The previous example illustrates a relationship between indicators and performance goals, such as revenue. In addition, balanced scorecards include measurements and relationships at all three levels: strategic, tactical, and operational. Balanced scorecards provide a much more active management tool than traditional financial reports.

Another important and growing trend is the BI "center of excellence." These centers of excellence combine staff with outside resources that have expertise in business, analytic tools and techniques, and IT. The business specialists are crucial to aligning a BI system's information and analysis with business goals and strategies. The analytic experts provide the intellectual horsepower to establish the right kinds of data gathering, analysis, and modeling. And the IT specialists implement the necessary infrastructure for data acquisition, storage, and access. An integrated approach has proved more successful than cases in which the IT organization has sole responsibility for setting up BI.

Basic Steps and Tips

When you embark on a major BI project, be sure to lay the groundwork properly before you start any implementation. The most important principle is to align BI projects with business goals. Here are some crucial steps:

• Identify the key decisions, who makes them, and when the decisions are made.
• Identify the right information that BI must provide for these key decisions.
• Identify the right time when this information must be available.
• Identify the right way to present this information.
• Work backward to the data required, including how it needs to be stored and refreshed (if not realtime).
• Identify the best tools to manage the data, provide the analysis, and present the information.

If this is your first major BI project, start with a focused project that has a well-understood ROI. Most initial projects should be ones that several other enterprises similar to your own have done, and don't pick an area that is politically highly charged.

As I described, identify a "team of excellence" within the organization to support the project. For an initial project, this team can be an ad hoc group; as you continue to expand BI, you may want to establish a permanent center of excellence.

Use proven platforms (e.g., the System i) and tools. BI has expanded well beyond basic reports, and few organizations are well positioned to create completely homegrown solutions. You may also find that your first project benefits from consulting support, either from a tools vendor or an independent.

Deliver concrete results early and often, both to make sure that you're on the right track and to establish organizational confidence that the project will deliver real value. Anticipate change, and plan for adaptability. More than most types of IT-related projects, BI projects tend to be highly fluid because managers don't know exactly what they need until they start seeing some results.

Look for established patterns and templates from other organizations in your field or from organizations using the same tools as your enterprise. When you're just getting started, you'll find that many of the BI capabilities that your organization wants to implement have already been done. Of course, direct competitors are unlikely to share their most sophisticated solutions; companies similar to yours but not actual competitors may be more helpful. Most IT tool vendors also have a library you can draw on.

The advances in BI tools and techniques present many opportunities for enterprises to leverage their IT capabilities. With this basic guide, you're prepared to make your efforts in support of BI a real success.

Paul Conte is a SystemiNEWS senior technical editor.