Implementing a Data-Driven DSS
Organization-wide Information System development projects like Data-Driven DSS are subject to numerous constraints. Some of these constraints are based on available funding. A large data warehouse can cost US $ 2-3 million for software, hardware, staff development time, and training costs and can take 2-3 years to build. Others are a function of management's view of the role played by an Information Systems department and of management information and DSS requirements. Also, constraints may be imposed by corporate culture conflicts. Letís now identify some issues that must be confronted when implementing a data warehouse, OLAP system or other Data-Driven DSS.
The first point to remember is that a data store is not a static database. Instead, it will be supplemented regularly with historical data. Because the data store is a foundation of a modern DSS, the design and implementation of a sophisticated data store provides an infrastructure for company-wide decision support. The decision support infrastructure also includes hardware, software, people, and procedures. A data store is a critical component of a modern Data-Driven DSS, but it is not the only critical component. The structure of the data store and its implementation must be examined in the context of the entire DSS infrastructure.
The technical aspects of creating a new database must be addressed. The new Data-Driven DSS must provide required analysis capabilities with acceptable query performance and the DSS must support the data analysis needs of decision-makers.
Traditional database design procedures must be adapted to fit the requirements of building a large DSS data store. Data is derived from transaction databases so a DSS designer must understand the transaction database designs. It is difficult to produce good DSS data when transaction databases are of poor quality or are inaccurate. So how should a data warehouse or Data-Driven DSS be developed?
Figure 7.3 5 Step Decision-Oriented Design and Development Process
A General Design and Development Process
Various consultants have customized their Data-Driven DSS development processes. Chapter 4 discussed the two general approaches called Systems Development Life Cycle and Rapid Prototyping. For small projects like a Data Mart, one can use Rapid Prototyping. For large projects, the following steps based on a typical data warehouse development process (see the textbook by Professors Rob and Coronell, 1997) are appropriate. This decision-oriented design and development process includes 5 steps (see Figure 7.3):
The first step is Initial Data Gathering or Diagnosis. This step involves identifying and interviewing key future DSS users, defining the main subjects of the DSS, identifying the transaction data model, defining ownership of data, assessing frequency of use and updates, defining end user interface requirements and defining any outputs and representations. The emphasis on decision-makers and decisions should be maintained in subsequent steps.
The second step is Designing and Mapping the Data Store. In a relational DBMS environment the first step is to design the Star Schema and identify facts, dimensions, attributes. Then one creates Star Schema diagrams, attribute hierarchies and aggregation levels. These conceptual models then need to be mapped to relational tables. In a multi-dimensional database environment the key variables and dimensions need to be defined. The data store houses the relevant DSS data.
The third step is Loading and Testing Data. Creating the DSS database involves preparing to load data, defining initial data to load, and defining update processes. Then analysts define transformations of the transaction and any external data, map from the operational transaction data, integrate and transform the data. Next, analysts load, index and validate the data, and finally verify metadata and data cubes or Star Schemas.
The fourth step is Building and Testing the Data-Driven DSS. Analysts need to create menus, develop output formats, build anticipated queries, test interfaces and results, optimize for speed and accuracy, engage in end user prototyping and testing, and provide end user training in a development environment. Decision-makers need to be heavily involved in building and testing the new Data-Driven DSS.
The final step is Rollout and Feedback. This step involves actually deploying the DSS, providing additional training, getting user feedback, maintaining the system, and in many cases expanding and improving the DSS. One hopes the DSS improves decision-making and benefits the company and decision-makers.
The above five step development process needs to be altered in some important ways when one is building an Executive Information System.