Decision support software

This article appears in the issue April 1998 [Volume 7,Iissue 5]

   Bookmark and Share

Intense competition and endless data make its future bright

As the discussion continues on knowledge management--what it is, how it differs from information management and so on, it's easy to forget why we want to manage knowledge in the first place. The central reason for managing knowledge is to make better decisions. Better decisions are those that have been arrived at using a valid process and that have a positive outcome. Decision support software (DSS) assists with structuring, accessing and analyzing data to support rational and timely decision making. It is rapidly becoming one of the boom niches of business intelligence (BI) software.

Three types of decision support software are discussed here:

  • online analytical processing (OLAP) software, which retrieves and analyzes structured data in response to user queries;

  • data mining software, which takes a more proactive approach by suggesting relationships among data elements;

  • group decision software, which structures and organizes input from both humans and computers to facilitate collaborative decision making.

Expert systems, also a type of decision support software, are not addressed here. They are often rule-based systems that incorporate technical expertise to provide a solution to a relatively specific question, such as whether an individual qualifies for a loan.

Like many aspects of knowledge management, the field of decision support software is fraught with definitional peril. The distinctions between the various types of decision support software are sometimes murky, and with good reason. Most of them have their roots in more than one field, and the boundaries can be fuzzy. For example, the query and reporting end of the BI spectrum clearly emerged from database technology, but data mining has a strong artificial intelligence component, and risk analysis is based in statistics. Some products combine elements of each.

Ten or more years ago when artificial intelligence was just becoming a mainstream concept, the belief was strongly held that standalone "expert systems" could provide answers to a wide range of technical, financial, medical and many other questions. Much money was spent on developing elaborate software products and cajoling them into simulating human thought, only to find that the expert system would be stumped by the most obvious (to humans) inconsistency. In a rule-based expert system, it was virtually impossible to predict and design for every contingency, and neural network-based systems were too esoteric to find their way into practical applications. Moreover, computing power was not up to the task at that time, particularly for any desktop applications of expert systems. Memory was small, and computing speed was painfully slow by today's standards.

By scaling back the initial ambitious goals and waiting for computer power to catch up, several branches of artificial intelligence have found solid niches. Artificial intelligence is not so much a field in itself now as an integral part of many other fields. No longer hoping to provide all the answers in a self-contained system, decision support software brings to the fore the information needed to facilitate decision making, and structures it into a meaningful presentation.

Most decision support products initially evolved from academic research, and many non-commercialized products have been developed. The transition from the lab to the market has been smoother in some cases than in others. Now that the field has matured somewhat, the larger software companies are beginning to take an interest, and some are developing DSS products in-house. For the less established products, the "merge or die" pattern is likely to hold.

OLAP and data mining software are relatively small markets at present, but are growing fast. Dataquest (San Jose), which uses the term "business intelligence" interchangeably with OLAP, reports that for this category, licensed revenues in 1996 approached $1 billion, up 25% from 1995. Data mining, though a much smaller market at about $20 million, was up 81%. That increase is partly accounted for by identification of new products. Said Dataquest analyst Peggy O'Neill, "OLAP accounts for the largest part of the market by far, but data mining is a fast growing component."

Group decision ware, although clearly in the category of decision support software, is not included in Dataquest's estimates of the market for business intelligence products. Developers of group decision products, however, are united in their conviction that the potential for this software to maximize the value of employee knowledge and expedite sound decisions is only beginning to be tapped.

Risk analysis, which falls into the category of analytic software, is part of a market that was assessed at $500 million in 1996 and is expected to reach $2.5 billion by 2001.

Market analysts classify the software products in different ways, but most seem to converge on a growth rate of around 30% per year.

We have the data--now what does it mean?

Over the past decade, organizations have stored vast amounts of data in repositories ranging from mainframes to desktop hard drives. More recently, the data has been organized into data warehouses. But making constructive use of the information has not always been easy. Most of the potential users did not have the database skills to effectively access and analyze the data.

One set of tools that aims to simplify data access is a suite of products from Brio Technology ("As organizations have been downsized and their structure flattened, it has become increasingly important to empower employees with the ability to access information directly," said Brio VP Nick Halsey. "We see the work environment changing from one in which 10% of the employees used query tools 100% of the time to one in which 100% of the employees use the tools 10% of the time." It is no longer necessary to be a database expert to make use of information in data warehouses.

BrioQuery supports ad hoc query, quick data sorts and other capabilities such as "drill everywhere" functionality. It can create 3-D charts and perform local calculations. Thus it is a tool for exploring and analyzing existing data to support decision making, but users need not be experts in database querying to do their analyses. Brio extends its OLAP capabilities to the Web with Brio.Insight, which has the same analytic capability as BrioQuery. "We were the first to offer full OLAP functionality on the Web," said Halsey. "The server handles not only the connection to the database, but also security that would not otherwise be provided by the browser."

From OLAP to data mining

Cognos (:// is recognized as a market leader in business intelligence software, with products that offer both OLAP and data mining capabilities. Cognos' OLAP product, PowerPlay, takes a somewhat different approach from that of BrioQuery in providing an aggregated view of the data first, with the ability to access progressively more detailed information.

"We have been using Cognos for three years," said David Bruce, manager of database systems for Random House (New York). "After looking at the aggregated data, we can easily drill down to look at sales data for a selected region right down to the zip code." Typical explorations include looking at sales data for particular groupings of books (referred to in the business as "imprints"), monitoring the effectiveness of promotional campaigns, and looking at patterns of book returns. "We needed an application that could handle 2 GB of raw data, because we analyze not just in-house data but external data from bookstores and other sources," said Bruce. "Cognos was able to provide this capability at a reasonable cost." Like Brio, PowerPlay has a Web-enabled version that provides full OLAP functionality.

For data mining, Cognos' Scenario provides a level of analytical capability that OLAP tools do not. "OLAP is about asking a question and getting an answer," said Tom Camps, Cognos director of product marketing. "Data mining is about looking at a data space and having the software tell you what is important." For example, Scenario could be used to perform a statistical analysis to determine what the key elements are in revenues, without having to suggest the factors in advance. Data mining software has a lot more built-in intelligence, including sophisticated neural network software. Version 2.0, announced in February, integrates seamlessly with Impromptu (another Cognos product) and PowerPlay.

Another company that provides a suite of OLAP and data mining tools is Business Objects ( Its BusinessObjects query tool can be used to perform manual data exploration, and BusinessMiner can then be used for automated data analysis. A model based on the data is built, and then a set of rules about the data is generated. For example, BusinessMiner can develop rules regarding customer groups that can be used for managing customer relationships or for targeted marketing. BusinessObjects' WebIntelligence product allows for enterprisewide querying of the Web.

Naturally the database companies are not content to sit still while other companies develop software that analyzes their rows and columns. Database powerhouse Oracle (, which is the leading producer of relational database software, has a full line of data warehousing products and several OLAP products. The Express series, including Express Server, Express Objects and Express Analyzer, is geared toward multidimensional analysis, similar to Cognos' PowerPlay. A newer OLAP product, Discoverer 3.0, is directed toward relational databases.

"We think that ease of use will be a key factor in determining market success," said Dr. Anna Wichansky, manager of Oracle's usability engineering laboratory. "Today's employees are not able to spend a week learning a new software package. They have to be up and running almost immediately." The wizard-driven query builder and the ease of data manipulation once the results of the query are presented are two of Discoverer's strengths. In an independent study, users were able to complete a test suite of tasks significantly faster with Discoverer than with BusinessObjects or PowerPlay. Users also were enthusiastic about the online help and tutorial.

Getting to a consensus

Of the many types of information resident in a company, human knowledge has been the most challenging to extract and use effectively. Getting last year's sales figures is a snap compared to mobilizing the unstructured, undocumented, but crucially important expertise contained in the staff of any enterprise. One category of software, group decision process software, aims to foster the process of applying and coordinating input from groups to reach a decision that is demonstrably sound and broadly supported.

Team Expert Choice, a decision support software product from Expert Choice ( be used either as a tool for group decision making or as a shell to build an expert system. Based on matrix algebra, the product supports the phases that occur naturally in moving toward a decision, and provides an environment in which all the factors can be visualized and displayed graphically. In the first phase, brainstorming, the key issues are identified. Second, those issues are structured. Finally, the evaluation and choice phase integrates the input into a format that presents the conclusions and allows for "what if" modifications.

In one case, a company had been trying over an extended period of time to select a vendor for a large-scale project, but the decision had become bogged down by political and technical dissent. Scitor Corporation, an engineering and management consulting firm (Sunnyvale, CA), used Expert Choice to lead the factions to a creditable and documentable decision.

Part of the initial problem was that too many decision criteria were included, which made the decision process impossible. Though the structuring process of Expert Choice, a manageable number of criteria were selected. Then a diverse group of staff members began assigning weightings to each of the criteria, and evaluating each vendor across the criteria. Despite significant controversy going into the process, a decision was reached within six weeks. During the process, the decision was tested by modifying the criteria ("What if we gave a weighting of 30% instead of 20% to this factor?") or the evaluations of the options ("What if company A was ranked higher on this factor?"). The results of the testing were surprising to the client, but not to Scitor. "The decisions reached through use of Expert Choice are very robust," said Scitor's Curt Vonder Reith. "Everyone involved perceived the final result as the legitimate conclusion of a valid process, and they were able to support the decision."

The graphical interface provides a highly visual presentation both of the results and of the options for modification. For example, a line representing weightings can be "dragged and stretched" to reflect a new weighting, and the new result can be seen immediately. Aside from the practical contribution of organizing information, the tool also has a value in keeping discussion focused on the issue. Clearly, the product is most useful in the hands of a skilled facilitator. "Although we wish that clients would approach us early in the decision process," said Vonder Reith, "usually they don't seek help until they have run into trouble. At that point, the value of Expert Choice becomes particularly evident."

In an application at Amoco (Chicago), Expert Choice was used in assessing the costs and benefits of potential new strategies for Amoco's Controller's Department. "At first, people felt that a decision already had been made," said T.J. Gallagher, a staff director in the Controller's Department. "But as the process went on, everyone came to feel that their input had been factored into the decision." Gallagher cited the team building value of using Expert Choice. "Everyone could see the factors that went into the decision and what variables were most important," he said. "It was definitely a participatory process. In a sense, the software functioned as an internal consultant."

Kumar Nochur, president of Vidya Technologies ( that while workflow products have helped organizations complete tasks more efficiently, they have not made organizations function more effectively. Efficiency is doing things right, and effectiveness is doing the right things. Vidya's ThoughtFlow software is designed to help address some of the most basic questions a company faces: What are the organization's goals, are they the right ones, and are plans and actions aligned with goals and strategies to achieve desired results, i.e. improve effectiveness?

ThoughtFlow is designed around a flowchart metaphor (in Release 1.4 for Windows) and a tree outline metaphor in the intranet version now under development. When the software opens up, it presents default "templates" for defining goals and strategies, aligning plans and actions, solving problems and making decisions. The elements on the template are represented by a different physical shape or icon to create a shared visual language in the minds of users. The templates can be modified and filled to create a map that reflects the specifics of concern to the company. The software allows an individual or group to visualize abstract topics. It is useful in providing a big picture or overview of the logical flow for decision making. Elements on the map can be linked either to other ThoughtFlow maps or to applications such as spreadsheets. ThoughtFlow can thus be used as a mechanism for presenting more detailed information that can influence decisions. Data relating to each decision element, such as priority, status and dates, can be defined for search and retrieval, task assignments and monitoring to see if goals are being achieved.

Typical applications for ThoughtFlow are strategic planning, decision support and product development. Said Vidya founder and President Kumar Nochur, "We are providing the big picture perspective. ThoughtFlow has been used in high-tech companies to create an audit that shows gaps in strategic planning and suggests some ways of filling in the gaps." Other types of users include people in product development teams working on projects at different sites who share information over a network. Thus, ThoughtFlow can be used as a collaborative product for a wide range of group processes within an enterprise; a Web-enabled version is also under development.

Enabling software

Two other categories of software deserve mention in connection with decision support software. One is the enabling software that puts data where OLAP tools can get to it. Those include data warehousing products, a category in themselves, and data moving products. OmniEnterprise from Praxis International ( is an example of "data moving" software that sets the stage for OLAP by allowing data sharing across different database environments. That requirement can come about in a variety of ways. For example, departmental databases that grew up over time may need to be merged or accessed from other departments. Another case is when a firm is acquired by another firm that uses different database software.

In either case, there is a need to access data across heterogeneous environments. "People doing OLAP don't want to see all the data--just what's relevant to their analysis," said Praxis Marketing VP Randy Corke "and it has to be there when they need it." OmniEnterprise can be set up to move data at specific times, and it can also move just portions of the database into the data warehouse. OmniEnterprise Version 2.3 supports long data types such as documents and images; while not discussed here, information contained in documents can be key to decision making.

Another category that is closely linked with decision support software is risk analysis software. Crystal Ball from Decisioneering (, for example, is an Excel add-on that calculates the probability of attaining a goal based on user input. A typical application is in financial planning. In an Excel model, the predicted growth rate could be input and a predicted size of the investment calculated, but probabilities would not automatically be calculated. With Crystal Ball, the user can select the probability ("I want to be 95% certain of attaining this goal") and then see how the investment requirements would change.

"Risk analysis and intelligent forecasting will be major growth drivers for OLAP vendors as they seek to assist their customers in the next phase of decision making," said Decisioneering's Michael Reagan, VP of sales and marketing. "Companies are making strategic decisions but there is a lot of uncertainty associated with these decisions that is not built into the models. We see huge potential in this area." Reagan notes that OLAP combined with risk assessment is a powerful combination because it allows for a higher level of sophistication in the analysis. Decisioneering, which has already formed OEM arrangements with several other firms to integrate Crystal Ball into their software products (including banking software and a decision software product), is presently in partnering discussions with several leading OLAP and enterprise applications vendors.

Figures of merit

In the decision support arena, 80% is a popular number. For example, in data mining, 80% of the information is purportedly found in 20% of the data, so selecting the right data is important. Otherwise, the data-crunching task becomes prohibitive. Moreover, 80% of the time and effort in data mining is just getting the data ready to mine. That means cleaning up the data and warehousing it, sometimes from heterogeneous environments. Whatever the exact numbers, it is clear that planning and preparation are crucial if the investment in OLAP, data mining, or other decision support effort is to pay off.

Costs for decision support software and its enabling technologies vary widely. Cognos' Impromptu, PowerPlay and Scenario are $695 each. Users who want the Web-enabled version of PowerPlay will pay a premium, however: $40,000 for a 100-user Windows NT version and $75,000 for a 200-user Unix version.

For the group decision process software, Vidya's ThoughtfFlow is $299 for a standalone version and $2,995 for a six-user version; Expert Choice is about $12,000 for an eight-person version and $1,000 for a single-platform version. Each one needs to be evaluated on how well it matches up to the issue at hand. Crystal Ball, the Excel add-on, is $495, and the Praxis data moving software, OmniEnterprise, ranges from $2,500 to $90,000, depending on configuration (supporting the claim that much of the cost in data mining is in the preparation phase).

For the future

Vendors will continue to integrate capabilities to simplify the overall process of data warehousing, OLAP, data mining and risk analysis. Group decision software developers will continue to expand linkages to quantitative data. Boundaries between the different types of products will continue to shift. In addition, vendors will work hard to make the software easier to use, knowing that interest will quickly trail off if users have to struggle with the software. However, the incentives are strong for both developers and users to make the market work. Given the competitive environment in which companies find themselves and the great amounts of data they must contend with, decision support software has a bright future.

Search KMWorld