Oops! The input is malformed!
Originally published 27 August 2008
When we began our careers in information technology, the term decision support system (DSS)1 was often used to describe the software used for querying, reporting on and analyzing operational data. At that time, most processing was done on mainframe computers, and both operational and DSS applications shared the resources of the mainframe operating environment. DSS use by business users, however, was limited by the performance impact of DSS processing on operational applications and by the complexity of the DSS environment.
DSS performance and complexity issues, and the need to keep historical data, led to the development of the data warehouse and business intelligence (BI) applications. The last decade has seen a steady increase in the use of both data warehousing and business intelligence.
Most of the BI applications developed to date have been targeted at senior managers and business analysts who are interested in monitoring the overall performance of the business or a specific business unit. The need, however, to use business intelligence to make more timely decisions, monitor and optimize daily business processes, and to deploy business intelligence to a broader user audience has led to a wide range of new BI techniques, technologies and vendors that extend the traditional strategic and tactical BI environment.
To support so-called operational BI, data warehouses are now being updated more frequently to support daily and intra-day business decisions; and when close to real time decision making is required, BI processing is being tightly integrated with operational business processes.
To improve usability for less experienced users, vendors are integrating their products with office productivity suites such as Microsoft Office and taking advantage of Web 2.0 technologies such as rich Internet applications. Vendors are also moving toward supporting a software-as-a-service (SaaS) deployment model and web-based applications such as Salesforce.com and Google Apps. BI connectivity to collaborative and business content systems is improving information and expertise sharing, and is also enabling the growing use of search tools and unstructured data in BI applications.
Most BI applications still provide strategic and tactical BI and access data managed in a data warehouse. The tight connection between business intelligence and the data warehouse has led to the assumption that data must be maintained in a data warehouse before it can be used for business intelligence. This assumption is wrong. There are an increasing number of BI applications that do not employ a data warehouse, either because there is no need to store the data in a data warehouse, or because it is not practical or cost effective to do so.
The illusionary tight connection between business intelligence and data warehousing is causing people to ask whether the term business intelligence has outlived its usefulness. Some people have come full circle and feel that perhaps we should return to using decision support system instead. Whereas we support this position, from a marketing perspective such a change would be difficult because business intelligence is firmly entrenched in the marketplace and decision support is considered old fashioned. As a compromise, we have started to use the term decision intelligence instead. You could think of decision intelligence as being DSS 2.0!
Figure 1 illustrates how decision intelligence brings together traditional and operational BI, operational processing, and the collaborative and business content environments.
Figure 1: An Architecture for Decision Intelligence
In traditional BI (Figure 1 center), applications query, report on and analyze historical data stored in a data warehouse, and produce strategic and tactical analytics. These analytics typically provide a set of data metrics or indicators that measure the actual business performance of either a group of related business areas or a single business unit. A well known architecture that supports traditional business intelligence is the Corporate Information Factory (CIF).
Operational BI enables more timely decision making. Traditional BI applications can support operational BI by updating the data warehouse more frequently and/or by directly accessing operational transaction and master data. As already mentioned, it may not always be necessary or practical to store the operational data in a data warehouse before it is analyzed. This is especially the case when close to real-time analysis is required and/or business transaction and event volumes are very high. Financial industry algorithmic trading and web analytics are good examples of applications here. In these cases, the traditional store data and analyze it model is replaced by an analyze data and store results model. This, of course, has implications for both data integration and data quality. This aspect of decision intelligence will be covered in a future article.
To support high data volumes and close to real time decision making, the operational BI processing is integrated into the operational environment as shown on the left-hand side of Figure 1. One key characteristic of operational BI is that it is process centric, unlike traditional BI, which is primarily data centric. The integration of operational BI into operation processes can take two forms. The first is where the operational process sends a request to a BI or decision service to obtain information (customer lifetime value score from a data warehouse, for example) or for a recommendation (reject a credit card transaction as fraudulent). The second form is where operational processes and their underlying activities pass event information (point-of-sale or RFID product scan, ATM action, telephone call, hardware alert) to a BI service that analyzes the information and displays event analytics on an operational dashboard and/or stores the analytics in a data warehouse. In some cases, the events may be stored in memory or persisted in an event store while being analyzed.
The collaborative and business content environment (right-hand side of Figure 1) is the third component involved in decision intelligence. Unstructured business content can come from the collaborative environment itself (office productivity suites, social computing tools), BI environment (reports, spreadsheets) and operational processes (purchase orders, call center logs). All of this content can provide valuable insight for decision making. For analysis purposes, unstructured business content can be converted into a structured or semi-structured format (typically XML) and stored in a data warehouse for subsequent analysis, or can be processed directly by content analysis applications (such as a text mining tool, for example) to create content analytics.
Another important aspect of the collaborative environment is that it enables business users to share information and expertise. This aspect of collaboration is increasing with the growing use of social computing (blogs or social networks, for example) in organizations.
Figure 1 shows how the decision intelligence environment brings together three types of business intelligence (process, data and content) and enables the sharing of information and expertise. The objective of decision intelligence is to improve decision making throughout the enterprise, regardless of employee position or role. Aiding in this decision process are the underlying business rules and policies that define corporate and best practices. One benefit of decision intelligence is that it enables organizations to better understand their business processes and the factors that affect business performance. This understanding improves the quality of business rules and policies and leads to better decisions and actions.
The consumers of decision intelligence may be operational processes (application-centric decision intelligence) or business users (user-centric decision intelligence). Most usage at present is user-centric decision intelligence that integrates the intelligence at the user interface level using technologies such as dashboards and portals. There will be a need as decision intelligence grows to provide a semantic layer between the business user and the underlying sources of decision intelligence to isolate the user from the actual physical decision intelligence environment.
Our motivation in creating the concept of decision intelligence is to break the misconception that you must have a tight connection between business intelligence and data warehousing. Data warehouses are not going to go away. They are very successful at producing both strategic and tactical business data intelligence and, in some cases, operational business process intelligence. There are, however, a growing number of solutions that don’t require a data warehouse. These solutions don’t replace data warehousing, but provide valuable additions to it. It is important to remember that data warehousing came about to overcome issues with earlier decision support systems, and modern technologies are removing the need to always store data first in a data warehouse before it can be used for decision making.
We would like to thank Scott Davis of Eyeris and the many vendors who have taken the time to review and provide valuable feedback on earlier versions of the decision intelligence concept.