Providers of decision support software will not succeed unless Information Resource Management (IRM) is successful at insuring the quality of the data being loaded into the information warehouse.
Access and quality are a 50-50 split. Both are needed become highly successful. This article analyzes the relationship between On-Line Analytical Processing (OLAP) and IRM from an OLAP vendor’s
perspective. I start with these questions:
- What has contributed to IRM’s struggled in the past?
- What is the value-added of IRM going forward?
- What is the scope of IRM for decision support?
What has contributed to IRM’s struggles in the past?
Many of us who have been involved in IRM initiatives have struggled to succeed. One reason is that our focus has been on the implementation of meta data repositories to support transaction/legacy
systems. Repositories help us document and manage legacy/source systems and are valuable for understanding and improving those transaction systems. However, one of the first functions eliminated
during the downsizing and business process re-engineering of the past decade was that of IRM. The documentation and standardization initiatives appeared to be valuable, but from the perspective of
each individual application there was little value added and business owners were not going to absorb the costs for integrating and standardizing data. There was no business champion and no
business need to support the “requirement” to integrate information across applications because the transaction applications are self-sufficient. Suffering from lack of a business need or
champion, IT has been unable to build a solid business case for the IRM function.
Enter Decision Support.
What is the Value-Added of IRM going forward?
The competitive advantage provided by improved decision-making capabilities (through Data Warehouses / DSS / …) has become the business driver for the existence of IRM functions, roles and
responsibilities. Decision support, whether we are talking about semi-structured operational decision processes or unstructured tactical and strategic analyses via OLAP, forces corporations to
address data integration and quality issues IRM has been trying to accomplish for years.
A key deliverable:
One key to the success of decision support lies in the elimination of the word ‘doubt’ from the user vocabulary. Clients admit that an inordinate amount of their research and analysis
time is spent eliminating doubt about the information they are analyzing. In one example, a client estimated that 80% of her research efforts are spent eliminating doubt rather than concentrating
her efforts on supporting the decision to be made.
Decision support initiatives will not survive unless this doubt factor is drastically reduced. Ideally, the doubt factor should be eliminated, allowing 100% concentration on solving problems and
exploiting opportunities. Accomplishment of this goal will allow users to concentrate on the decision processes, not the quality of the data.
To eliminate doubt and insure high-quality data and consistency in decision making, companies are forced to standardize and integrate shared information. Sharing, by itself, will not solve data
quality problems while producing decision support systems. From a decision support perspective, data warehouses provide as much motivation for IRM as transaction systems (i.e. they operate in their
own world). The motivation for companies to re-implement IRM doesn’t appear until they have a critical need to eliminate doubt from one or both of the following decision processes:
- operational decision processes (e.g. customer service representatives; account managers); or
- tactical and strategic decision processes above the “mart” level (e.g. divisions above departments; the evolution-to-enterprise expansion of warehouse applications, forcing the sharing of
dimensional and fact data).
What is the Scope of IRM for Decision Support?
Those of you who have followed the birth, burial, and current resurrection of “Decision Support” may recognize the three-legged stool analogy for this topic. Decision Support has been discussed
as the marriage of three components: data, dialogue and mathematical models. These components have been independently addressed by three disciplines: information science; cognitive psychology; and
mathematical modelers (management science/operations research/statistical analysis). In the past, DA has focused almost entirely on the design and management of the data component. Mathematical
modelers have researched and discussed the management of the math models, but not the data.
IRM is forced, by business need, to address the management of both the data and the mathematical models. The following thoughts are offered to expand the readers thinking about the management of
the mathematical models which I will characterize in three types:
- Calculations
- Business rules
- Complex algorithms (not covered in this article)
Management of Calculations
Many calculations used in business today are buried in macros executing in spreadsheets and are therefore: undocumented, non-standardized and virtually invisible to the organization. Companies who
have OLAP applications in production against a warehouse have already been confronted by the need to manage calculations. OLAP provides both the need to standardize and the opportunity to
manage calculations because it elevates the visibility of the calculations. Because of the power of calculations and the potential for misunderstanding, standards and consistency of
calculations may be even more important than they are for data that is extracted, transformed and loaded into the warehouse from source systems.
Management of Business Rules
It is interesting to see how important business rule management is becoming to the IRM community. Business rules, from a decision support perspective, are used by companies both to monitor and
control the efficiency of their operational processes and to govern the empowerment of their employees to make decisions. It’s this latter use that caught my attention because no operational
decision-making is done without constraints. The values and relationships contained in the constraints which bound operational decision-making authority are a form of business rules, which must be
monitored, controlled, and improved.
The tracking and improvement of both transaction process efficiency and operational decision-making effectiveness is one of the primary targets of OLAP decision support.
Management of Complex Algorithms
Consider where evolutionary design of decision support systems will take us in the coming years. Once end-users trust their warehouse, we will be allowed to unleash the power of mathematical models
(e.g. data mining, statistical analysis, optimization techniques) against that same data. The mathematical modeling community has addressed ‘model management’ for years, but this is not
the same repository-type management with which the typical IRMer is familiar. It has to do with the monitoring, control and improvement of the components (e.g. coefficients, exponents) within the
math models and the documentation of the underlying assumptions contained within the model. Because of the power of the mathematical models used in decision support, this topic will become at least
as important to IRMers as data management is today.
Hitch your wagon to a star.
Why this title? Those of you familiar with OLAP’s world of dimensions and facts have already figured it out. For those of you who aren’t, dimensions and facts are what make up the star
schema. Decision support, especially the OLAP aspect of decision support, provides recognizable, immediate corporate value-added that is as dependent on IRM as it is on the use of OLAP access.
Therefore, the DA industry should not only look to the meta data repository of the source systems feeding the warehouse, but more importantly look to the enterprise warehouse and hitch their wagon
to the OLAP star schema.
