Business Intelligence/One frame at a time

“	...if one carefully examines explanations produced by analysts, a number of fundamental similarities emerge. Explanations produced by particular analysts display quite regular, predictable features. This predictability suggests a substructure. These regularities reflect an analyst's assumptions about the character of puzzles, the categories in which problems should be considered, the types of evidence that are relevant, and the determinants of occurrences. The first proposition is that clusters of such related assumptions constitute basic frames of reference (our emphasis) or conceptual models in terms of which analysts both ask and answer the questions: What happened? Why did the event happen? What will happen? Such assumptions are central to the activities of explanation and prediction, for in attempting to explain a particular event, the analyst cannot simply describe the full state of the world leading up to that event. The logic of explanation requires that he single out the relevant, important determinants of the occurrence. Moreover, as the logic of prediction underscores, the analyst must summarize the various determinants as they bear on the event in question. Conceptual models both fix the mesh of the nets that the analyst drags through the material in order to explain a particular action or decision and direct him to cast his net in select ponds, at certain depths, in order to catch the fish he is after.	”
— Allison 1969

Business Intelligence System: Architecture and Infrastructure

Why build a BI system one frame at a time? This chapter will answer this question by first defining a frame as an information system. The next section will discuss a few approaches to building BI systems. After discussing other approaches the chapter explains the framework approach to build a BI system. The next section notes that this approach is compatible with other approaches. Finally, the chapter ends by demonstrating the benefits of the step-by-step approach to building a BI system offered by the framework.

How do decision makers view the world?

Herbert Simon wrote on the cognitive aspects of decision making. For him a decision involves a rational choice selected from a number of alternatives towards the achievement of a goal. Rational choice means the decision maker can list the outcomes and order them in terms of preference. The individual then chooses the action he/she believes will deliver the preferred consequence or set of all the possible consequences. Formally, Simon argued that decision making is a process that can be divided into three required steps (1976):

1. Identify and list all alternatives
2. Determination all the consequences resulting from each of the alternatives
3. Compare the accuracy and efficiency of each of these sets of consequences

This describes an ideal situation. However, there are many reasons that decision makers cannot fully engage all three steps. Simon argues that the rationality of individuals is limited by such factors as the information they have, the cognitive limitations of their minds, and the finite amount of time they have to make decisions (to name just a few). The key argument is that the decision maker rarely has the time necessary to identify all the possible choices, determine the outcome of each choice, collect information on the relationship between choices and outcomes, and then analyze the situation to determine how each choice relates to each outcome.

To explain this limitation he introduces the concept of bounded rationality. Bounded rationality means that decision makers would like to engage in all three steps listed above but limitations beyond their control require them to make shortcuts. Instead of engaging fully in all these three steps the decision maker satisfices. Simon defined satisficing as the situation where a decision maker focuses on adequate solutions rather than optimal solutions. In order to understand decision makers we need to first understand decision making.

Given bounded rationality how do decision makers satisfice? Decision makers (or organizations) use cognitive processes. Simon argued that

“	The human being striving for rationality and restricted within the limits of his knowledge has developed some working procedures that partially overcome these difficulties. These procedures consist in assuming that he can isolate from the rest of the world a closed system containing a limited number of variables and a limited range of consequences.	”
—Simon 1976

The decision maker or organization narrows down the number of variables using values. "Decisions can be complex admixtures of facts and values." [1] The fundamental assumption here is people and groups have values and experiences that shape how they make choices.

Graham Allison provides an important explanation of the internal influences on behavior by talking about reference frames and how these frames influence behavior.

“	...if one carefully examines explanations produced by analysts, a number of fundamental similarities emerge. Explanations produced by particular analysts display quite regular, predictable features. This predictability suggests a substructure. These regularities reflect an analyst's assumptions about the character of puzzles, the categories in which problems should be considered, the types of evidence that are relevant, and the determinants of occurrences. The first proposition is that clusters of such related assumptions constitute basic frames of reference (our emphasis) or conceptual models in terms of which analysts both ask and answer the questions: What happened? Why did the event happen? What will happen? Such assumptions are central to the activities of explanation and prediction, for in attempting to explain a particular event, the analyst cannot simply describe the full state of the world leading up to that event. The logic of explanation requires that he single out the relevant, important determinants of the occurrence. Moreover, as the logic of prediction underscores, the analyst must summarize the various determinants as they bear on the event in question. Conceptual models both fix the mesh of the nets that the analyst drags through the material in order to explain a particular action or decision and direct him to cast his net in select ponds, at certain depths, in order to catch the fish he is after.	”
— Allison 1969

The approach to building business intelligence systems presented in this book assumes that implicit conceptual models influence our thoughts and behavior. Given bounded rationality every individual must apply a method to filter the flood of information that bombards them daily. They do this using their frames of reference, used both to filter data and make sense of the data. What the decision maker sees and judges as important is based on information that passes through "conceptual lenses" that all individuals use to make sense of information. Decision makers frame policies in terms of what Allison explains as conceptual models that have significant consequences for the content of their thought (1969). In sum, what each decision maker sees and judges to be important is a function not only of the evidence but also of conceptual lenses through which we look at evidence.

At the heart of a Business Intelligence system is a process that filters and then organizes vast amounts of data into information and knowledge. Explicit strategic models that capture reference frames are an important part of a Business Intelligence system. The system filters out data the decision maker deems unnecessary and organizes useful data into information based on a frame of reference. Decision makers hold certain information and events as critical for gauging strategic and operational efficiency. Building a Business Intelligence system that applies this frame to incoming data helps the decision maker make sense of the world.

Capturing the decision maker's reference frame

We capture the strategy of decision makers by modeling their reference frame. The process for modeling strategy is based on developing strategy diagrams. We propose that the first step in framing a strategy is to develop strategy diagrams based on frames of reference. There are four types of strategy diagrams. The term strategy diagrams is used instead of strategy maps because Kaplan and Norton developed the Strategy Map. Further, Porter develops the Activity Diagram. To avoid confusion strategy diagrams refers to the collection of maps. Any one is called a strategy diagram. Strategy Map refers specifically to the approach to strategy developed by Kaplan and Norton. Note that a map and diagram are generally the same thing.

The strategy diagrams are as follows (authors in parentheses):

1. Activity Map - (Porter 1996)
2. Chatterjee Map - (Chatterjee 2005)
3. Strategy Map - (Kaplan and Norton 2001)
4. Causal Link Map - implicit or explicit part of the previous three maps

This paper asserts that the Kaplan and Norton's Strategy Map and Balanced Scorecards can be used to support two distinct management activities – management control and strategic control. The paper describes characteristics of Balanced Scorecards appropriate for each purpose, and suggests a framework to help select between them.

Approaches to building Business Intelligence Systems

In developing a Business Intelligence system there is a huge gap between the multitude of fragmented data sources and executives that need access to the information that these data sources can provide (see Barberg). Most approaches to building BI systems must address these problems. Further, the choice is even more extreme when building a BI system from scratch.

To this end there are two approaches to building BI systems. Each focuses on the link between data and decision makers.

• Bottom Up - Enterprise Data Warehouse Approach
• Top Down - Executive Dashboard Approach

Enterprise Data Warehouse Approach

This approach begins after the creation of an enterprise data warehouse. While the data warehouse is complete there are still users that do not have access to the data they need. Further, they are often exposed to data and reports that they do not want. Finally, many argue that the data warehouse is a resource to be tapped by a BI system (as easy as stacking Legos).

"Even if, after many months (or years) of effort, the enterprise data warehouse is a technical masterpiece, there still may be a big gap between the data warehouse and the end users. Many companies have spent many millions in building a solid data warehouse foundation only to have the business users fail to embrace it. Part of the gap was filled, but a bridge half-way across a wide river is of very limited value." [2]

"Vendors like SAP and Oracle are finally building in what users thought they were going to get when they invested millions in an ERP, which is great reporting and great business management tools. But the problem is that [the ERP vendors] are building from the bottom up. They built a data model for business applications, with a layer of analytic applications on top of it.

With Oracle's acquisition of Hyperion and SAP's acquisition of Business Objects, they both recognized that approach of building it from bottom up and adding it onto the ERP was not really getting onto the broader BI market."[3]

The bottom up approach is often driven by the notion of tapping the Enterprise data warehouse as a new resource without any thought regarding how to use this new resource to deliver information.

Top Down

A purely “top down” effort that focuses on delivering executive dashboards and scorecards can also prove to be a costly disappointment. There are many different approaches, ranging from manual spreadsheets of performance metrics to sophisticated Business Performance Management (BPM) software packages that promise to deliver key insights on a vast number of performance indicators. Simple scorecarding or dashboard packages may provide some executive-friendly briefing books, gauges and charts, but they often lack the ability to evolve into a richer analytical platform.[4]

Framework Approach to build a BI System

A framework approach to a BI system includes both an architecture and infrastructure for each frame and an architecture and infrastructure for the framework. These are outlined below.

Frame Architecture

This book focuses on building Business Intelligence systems from a framework approach. This is a top-down approach. The framework approach starts with building frames for individual decision makers. The frame is developed from strategy diagrams. The strategy diagrams model strategy from all the decision makers' combined reference frames. Each frame is a slice of the whole corporate strategy and corresponds to the frame of one decision maker (see Barberg).

To build the frame the Strategy Map (Norton and Kaplan's) is cascaded down to the department level in the form of a scorecard (see Barberg). This is the first step in building the frame architecture. This is because the frame architecture must be based on the overall company strategy. In fact, at the heart of the architecture for the frame is the decision maker's view of the strategy, cascaded from the strategy map.

This creates a connection between the overall company strategy and the individual frames. At the operational level this means that the frame focuses on the processes and activities outlines in the Strategy Map. The frame owner focuses on the operations within his/her control. The scorecard allows the owner to identify the important objectives that will bring the strategy into fruition.

Frame Infrastructure

To determine if the business unit is making progress towards achieving its objective the frame owner focuses on the KPI's and targets on the dashboard. The dashboard is at the heart of the decision maker's frame. The dashboard contains the most important information needed to determine if the business unit is making progress towards the achievement of its goals. The specific goals cascade down from the Strategy Map. The KPI's and targets in the dashboard instantiate the corporate strategy.

The reporting system has a hub and spoke architecture. The center of the reporting system is the dashboard. Each spoke is a report on an element of the dashboard, such as a report on a specific KPI. The frame owner can gain further information on any dashboard item by examining the associated report. The details helps the owner understand how his/her business unit is meeting its goals or why it is not.

In order to feed data into the reports the frame architecture includes a data mart. This data mart contains all of the data necessary to feed the reports and dashboard. The data mart can receive data from multiple sources but the data warehouse is the optimal feed.

Framework Architecture

Strategy maps allow for the creation of a control system that acts as the central nervous system for a BI system. A control system is a management tool used to intervene in organizational behavior when there is variations between actual and expected results (Cobbold and Lawrie 2000b).

Managing an organization requires:

1. Planning - a process of creating a statement of intent
2. Control - assuring that desired results are obtained by having a mechanism "by which execution against the plan can be controlled." (quote from Cobbold and Lawrie 2000b; also see Anthony 1965)

Before going into detail on a control system it is first necessary to define planning, control and control system.

Planning

Planning is the activity where the business leaders decide which goals are the most important and how much activities and resources to direct towards the business processes that will attain these goals. Leaders must make choices about which outcomes to pursue and how to achieve these outcomes.

Control defined as “assuring that desired results are obtained” (Anthony, 1965). mechanism by which execution against the plan can be controlled. a set of tools and skills that will support the communication and implementation of the decisions made (‘control’).

Control system

A control system combines both planning and control. It includes management methods of creating and disseminating plans and alterations in organizational behavior needed to respond to such variations between actual and expected/standard results.

For a control system to be effective Cobbold and Lawrie argue that it needs to be informed about the activities and results of organizational processes in a manner that allows a comparison to prior expectations or standards (2000b). Vickers writes that a control system needs “a means of comparing any state, actual or hypothetical with a standard” (1958). In BI systems these are measures and targets.

To ascertain whether or not the destination state is being obtained, the organization needs some form of feedback on activities being undertaken, and the outcomes arising from these activities. As noted above, without measurement it is difficult to track delivery of plans, and so a key task here is the selection of the right measures to inform managers about activity and outcome.

Finally, the framework develops with the development of frames. Frames can be linked horizontally or vertically. For instance, operational frames can feed into tactical frames. Tactical frames can feed into strategic frames. Further, horizontal integration allows knowledge to be passed between business silos. The BI framework is greater than the sum of its parts.

Finally, the framework approach allows for the connection of strategy and control through the firm by cascading scorecards. "The BSC also provides a consistent language and best practices for things like cascading the top scorecard down through the organization to make sure that the scorecard measures help align scarce resources to accomplishing the strategy (see Barberg)." A key feature of the framework approach presented in this book is the ability to align measures at each level of the organization with the strategy. This ensures that all resources go towards making the company as efficient as possible.

Bounded Rationality and the Control System

Cobbold and Lawrie argue that how the organization chooses measures and targets a key element to whether or not a good control system is effective (2000b). One method is to use outside experience or benchmarks. Another way to select these measures is based on ‘desired outcomes’ (Anthony, 1965, Kaplan and Norton, 1992; 1996). A refinement of this method focuses on the need for the organization to ensure that the control system adopts measures and targets based on desired outcomes but adds they should restrict "measure selection to those that are relevant to these outcomes" (Cobbold and Lawrie 2000b).

Why should they restrict measure selection? Based on principles of bounded rationality. Decision making based on bounded rationality assume that decision makers have access to limited information, cognitive limitations minds, and limited time in which to make decisions[5]. Gathering all information would cause overload. It would also make the organization slow to react to environmental changes or operational problems. Decision makers do not have the ability to consider all options and therefore must simplify the number of choices by reducing alternatives. In other words, they satisfice, meaning that they look for adequate rather than optimal solutions [6].

Given these limitations how do leaders make decisions? What the decision maker sees and judges as important is based on information that passes through "conceptual lenses" that all individuals use to make sense of information. The framework approach to building Business Intelligence systems assumes that implicit conceptual models influence our thoughts and behavior. Decision makers filter the flood of information that bombards them daily using their frames of reference. Frames not only filter data but are also used to make sense of the data. Decision makers frame policies in terms of what Allison explains as conceptual models that have significant consequences for the content of their thought (1969).

The framework approach to Business Intelligence (BI) creates an infrastructure that filters and then organizes vast amounts of data into information and knowledge. The filters are based on the decision makers frames of reference, captured in strategy diagrams. These explicit strategic models capture reference frames and therefore form the heart of this Business Intelligence system. The system filters out data the decision maker deems unnecessary and organizes useful data into information based on a frame of reference. Decision makers hold certain information and events as critical for gauging strategic and operational efficiency. Building a Business Intelligence system that applies this frame to incoming data helps the decision maker make sense of the world.

A control system must take into consideration the limitations of the organization. A framework approach, applying the principles of bounded rationality, satisficing and frame of references, allows the organization to both account for its weaknesses and take advantage of knowledge (strategy diagrams).

Framework Infrastructure

The framework infrastructure includes all the technology needed to implement a BI framework architecture. These include the data warehouse (which develops naturally from the data marts), technology to capture feedback (success towards completion of goals), tools for dashboarding and reporting system, etc.

This is a short introduction to the frame approach. However, it does provide a rough outline of the benefits of building a BI system, step-by-step, one frame at a time.

Framework approach and its compatibility with other approaches

Bill Barberg focuses on an approach that allows the firm to start small and build. This approach is a much more orderly method than would be the case if efforts were made without starting small. In building part of the BI system each builder benefits from the lessons learned by other builders rather than having to learn them through the school of hard knocks — which can be both expensive and time consuming. In building a part of the BI system, the data warehouse, he uses Ralph Kimball. Kimball's approach to data warehouse development focuses on a dimensional data mart architecture that facilitates the addition of data marts over time with a high degree of consistency. Similarly, Poe et al. argue that the first data mart is also a data warehouse with the ease of adding new data marts to the initial data warehouse (1997). The first data mart becomes the initial phase in a small scale data warehouse.

This is consistent with the framework approach in that each successive frame contains one data mart. As the number of frames grows it is possible to integrate these data marts into one central data warehouse. Further, in a company that already has a data warehouse, the framework approach allows the BI system architect to integrate the frames with the data warehouse.

The framework approach is a helpful methodology for building a Business Intelligence system because it addresses an important problem with building a BI system. Executives can become overwhelmed with measures, objectives, data and tools. The framework approach, with its application of the balanced scorecard methodology and step-by-step approach, provides a way to focus on the most important measures that drive the organization's strategy. Strategy Maps and Balanced Scorecards are an established tool for developing and disseminating strategy. Executives need only focus on helping the business analyst develop the strategy diagrams. They do not need to focus on helping develop the whole architecture. Further, the growth of the system is incremental. This piecemeal, top-down approach to building the BI system using a divide-and-conquer strategy makes the system easier to manage.

Conclusion

The application of a framework approach to building Business Intelligence systems aligns measures at all levels of the organization with the overall strategy by first developing strategy diagrams (culminating in the Strategy Map). The next step is to cascade the scorecards to each frame. This serves as the basis for the frame architecture. One the frame infrastructure is completed the business analyst turns his/her attention towards the BI framework architecture. This develops naturally from the Strategy Map and Balanced Scorecard tools. Then the framework infrastructure combines all the frames. Most importantly, the divide and conquer method allows the firm to start building a BI system without becoming overwhelmed with details.