Intelligent Enterprise

Better Insight for Business Decisions

RSS
Webcasts
Whitepapers
Subscribe
Home

June 30, 2003

A New Analytic Perspective

Decision makers throughout your enterprise need an IT architecture that serves their needs — rather than the other way around. Here is a view of emerging enterprise analytic systems that uses BI and analytic requirements as the point of origin

by Kemal A. Delic and Umeshwar Dayal

Continued from Page 1

Decision-Making: A Matter of Style

Perfect decisions require perfect decision makers — and timely support from perfect data and information sets. We know reality is slightly different. Real-world decision makers live in an imperfect world. To give structure to this kind of world and understand groupings of potential consumers of analytic artifacts, we can think about three generic groups of decision makers: executives, managers/analysts, and employees. We will briefly discuss each group's decision-making characteristics.

A major challenge for executive decision makers is dealing with huge volumes of weakly structured data sets and other imperfect information sets in a timely manner. Top executives should be able to operate across terabytes of raw enterprise data that is aggregated, abstracted, and transformed into higher-density artifacts (such as maps, simulations, and insights) that can be presented to them via dashboards and other summaries.

Managers and analysts will be exposed to analytic artifacts created from gigabytes of data before reaching decisions. Finally, lower-level employees also encounter a daily inflow of megabytes of multimedia data and information. Anecdotal evidence indicates that these three generic groups of decision makers make (respectively) intuitive, rational, and routine decisions (see Figure 1).

Enterprise Analytic Systems: Supporting Decisions

Enterprise analytic systems cover different domains, address varying value-at-risk entities, and require different analytic artifacts to improve and accelerate decisions. In practical terms, supporting decision-making may mean exposing decision makers to expectations, predictions, and forecasts so that they can rationalize their options. To outline possible consequences and envision scenarios, developers must consider users' expectations. For example, cause-and-effect models will produce predictions. Randomness is a key factor in forecasting, which makes the creation of large systems for modeling and forecasting outcomes and behaviors difficult and complex. We can consider each generic group of users more closely:

Executives are senior employees responsible for strategic decisions. Perhaps without knowing it, they digest a huge amount of data and information to make a strategic decision about how to improve corporate client profitability, for example. Executives also make intuitive decisions, and do so with input from the enterprise data and information domain, and with external information about market conditions as well as their views of partners and suppliers. The time horizon for their strategic decision-making is usually weeks or months.

Managers and analysts are employees with specialized skills responsible for tactical decision-making. To make decisions about how to optimize cash flow, for example, they digest a modest amount of data and information. Characteristically rational decision makers, managers and analysts focus on expertise-related subjects involving relevant, but limited data and information sets. Their decision-making timeframe is typically days or weeks.

Employees making routine decisions about how to optimize employee communication costs, for example, digest a limited amount of data and information related to their domain. Employee decisions typically have a timeframe of hours or days.

Different users of enterprise analytics use different artifacts. Senior executives and analysts typically require interactive content (sometimes called "active content"), while larger communities of employees require passive (or static) analytic content. Each generic group we've outlined also exhibits different usage patterns. Analysts typically need a powerful client/server setup; managers need browser access to analytics from a portable Web-connected machine; and lower-level employees might need analytics right on their desktops or accessible via temporarily connected PDAs, mobile phones, or other portable devices. These are important considerations for the design and architecture of analytic systems.

We can now outline what we propose as a stratified enterprise architecture arranged into different computing layers that support different needs and decision-making styles.