Research in psychology and decision sciences has challenged the assumption that people apply rational principles in making decisions that work to optimize expected results. This descriptive, empirical research has catalogued many of the biases that enter into decision making. Managers often seek to mitigate uncertainty by any means before committing to a decision.
One of the most famous examples of this line of research is the study of "anchoring" conducted by Daniel Kahneman and Amos Tversky in the 1970s. Through many experiments, they were able to show that managers make judgments by making adjustments from some initial value, even if that initial value is based on totally random information. (Professor Kahneman was awarded the 2002 Nobel Prize for Economics for his pioneering research.)
Decision-making processes need to be improved, but which processes are the best candidates for improvement?
Strategic vs. Operational Decision Making
There are two types of decisions - operational and strategic. Strategic decisions have broad organizational scope but are infrequent, while operational decisions have more restricted scope but repeat frequently. IDC research shows that progress is being made by capturing and automating decision-making processes for repeatable, operational decisions.
Here are some examples of operational decisions:
- Do we extend credit to this customer?
- Are these transactions evidence of fraud?
- How can we reroute this shipment to meet the promised delivery date?
- What book do we recommend to this customer?
- Should this supplier be on the approved list?
By contrast, here are examples of strategic decisions:
- Do we acquire company X or company Y?
- Do we target retail or energy companies as an added vertical?
- Is it time to discontinue a product line or to launch a new one?
Strategic decisions are important. However, because they are infrequent, there is little opportunity to apply lessons learned on an ongoing basis and to provide software-based automation to support such a process. That is not the case with operational decisions which are repeatable.
Here are examples of operational decision-making initiatives across a variety of industries:
- National Bank: Customer segment managers (formerly managers for individual products) meet every Monday, armed with the output of buyer behavior models, to decide which products to offer to each segment.
- Manufacturer of Electronic Devices and Components: In the R&D division, best practices are being captured to determine which lines of research to continue and which to halt - resulting in significant savings in R&D expenses.
- Hewlett-Packard: An analysis of PC warranty claims provides an early warning of product defects, enabling remedial actions in the manufacturing process and proactive communications to customers.
- Banco Espirito Santo: This company deployed an early warning system to detect actions leading to customer attrition. By using predictive modeling software, they were able to improve "at-risk" banking customer retention by 50 percent.
Traditional vs. Decision-Centric Business Intelligence
IDC has stressed the importance of the policy hub as the critical link between analytics and action in a closed-loop system. It is the point in a business process where decisions are made and where the results of the decision are communicated to the people and the operational systems that are impacted.
Traditional business intelligence (BI) concentrates on information access by and delivery to individuals. The classic issues are:
- How to make the underlying data available.
- How to format the data into reports or multidimensional cubes.
- How to deliver this information to knowledge-workers based on their roles and responsibilities.
Decision-centric business intelligence (DCBI) extends traditional BI in the following ways:
- Adding collaborative support on top of access to information by individuals.
- Supporting decision-making processes, especially for repeatable, operational decisions such as anti-money laundering, fraud detection and portfolio analysis.
- Filtering, monitoring and delivering information based on relevance, enhancing the insight of decision-makers and determining how to deliver this information to knowledge- workers based on their roles and responsibilities.
The Learning Gap Between Information Delivery and Decision Making
The disconnect between the delivery of information and the processes of decision making, as shown in Figure 1, comprises what IDC calls the "learning gap."
The learning gap operates in two respects, from the upper half of the process (information monitoring and delivery) to the lower half (decision making) and from decision making to information monitoring.
First, there is a disconnect from information monitoring and delivery to decision making. Traditional BI ends at the "deliver information" stage, paying insufficient attention to forming a problem statement, searching for candidate solutions ("Hypothesize") and evaluating the likely outcomes ("Model"). The danger is that a decision-maker without information relevant to the decision at hand is likely to rely exclusively on intuition - a notoriously unreliable practice. (See Hammond, Keeney and Raiffa, "The Hidden Traps in Decision Making," Harvard Business Review, September-October 1998.)
Second, there is a disconnect from decision making to information monitoring and delivery. If decisions are not captured, there is no opportunity to track information relevant to measuring whether or not the decisions were effective. For example, a demand planner in a high-tech company is responsible every month for forecasting demand for each of the next six months. The forecast is then provided as input for a scheduling system, yielding a schedule that is used by a contract manufacturer. When there are last minute adjustments after the schedule is committed, the contract manufacturer can assess penalties. There was no tracking and analysis of past forecasts for learning and improvement. Then, the high-tech company implemented an analytic application that captured each forecast and analyzed deviations from actuals over time. With this type of feedback, the planner was able to improve accuracy to a significant degree.
If decisions are captured, criteria can be established for prioritizing information in order to track the effectiveness of the decision. This has two advantages:
- Using decisions as filters for information monitoring and delivery helps an organization determine whether to stay the course or trigger a reexamination and restart of the decision-making process.
- The workflow for decision making indicates which individuals should be alerted when expected outcomes do not come to pass.
In other words, if you know what decisions are made and who made them, you can be far more precise in tracking relevant information and alerting those decision-makers in time to revisit a decision. Here is where decision-centric BI intersects with business activity monitoring (BAM).
For example, consider a BAM offering such as Celequest with a sophisticated rules capability to monitor real-time or streaming data for events that meet defined predefined criteria. When the condition is detected, the rules engine sends an alert to responsible people. There is great flexibility in the types of views and event conditions that can be defined as a rule.
Linking such a BAM product to DCBI (decision-centric BI) would mean driving the definitions of the monitoring rules from the meta data on the actual decisions that were made and the individuals responsible for them. As new decisions are made, event monitoring/alerting rules would be defined or revised.
A Look Ahead: Vertical-Specific, Event-Based, Decision-Centric Analytic Applications
The ability to capture, monitor and analyze decisions and their impact requires rich, higher order meta data constructs for defining a decision and related events. These constructs must then be mapped to the meta data for the underlying structured and unstructured data that document the decision. This meta data mapping should enable access methods for end users who can search or query by specifying the relevant events and decisions. This convergence of DCBI and BAM would close the learning gap in a truly automated way.
As organizations recognize the need not just to deliver information but also to address decision-making processes, analytic applications will become more decision-centric. We already see evidence of applications in specialized areas - notably financial services. Applications for anti-money laundering (Mantas) or fraud detection (Fair Isaac) are examples. Other vendors are providing industry templates. The most logical choice is to start with financial services, as have BlackPearl and Orenburg. Spotfire, on the other hand, is a decision-centric vendor that began with pharmaceutical R&D and oil exploration.
DCBI is more than a vision. With an initial foothold in financial services, decision-centric analytic applications are beginning to appear across industries and domains. These applications will incorporate predictive, collaborative and event-based capabilities to optimize, automate and track decision-making processes.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access