2. Literature review
2.1 Business intelligence
2.1.1 BI infrastructure and data warehouse
Dispositive data can be defined as data that support management in answering key strategic questions or that assists in guiding management to make relevant decisions. The storage of such data occurs in a DWH, a centralized data base. The idea is the storage of all strategic relevant information in a structured way. Therefore, the DWH system represents the infrastructure of a BI system because it provides important information for a company’s management and the requirements for data recording (Chamoni & Gluchowski, 2006). Further, a regular data warehouse tool offers capabilities to secure the environment (Mantha, Manthey, Valeyko, and Yonce, 2014). The data with which BI tasks are performed
often comes from different sources – typically from multiple operational data bases across departments within the organization, as well as from external data vendors. Different sources contain data of varying quality, use inconsistent representations, codes and formats, which have to be reconciled. Thus the problems of integrating and standardizing data in preparation for BI tasks can be rather challenging. Efficient data loading is imperative for BI (Chaudhuri, Dayal, & Narasayya, 2011). Further, current cost accounting systems are sophisticated but decentralized across departments. Often, half of employees’ time in analyzing a problem is taken up with just getting the relevant cost data together. With DWH this situation can be improved (Moore et al., 2012).
Simply put, operational systems provide a basis for data entry and the analytical DWH system allows for an organized and comprehensible output of huge amounts of data (Chamoni & Gluchowski, 2006). Gansor, Totok and Stock (2010) present the introduction of a DWH as the continuous development of the management supporting systems that is suitable for data management and integration. This DWH is a companywide system for data integration to control the company (Chamoni & Gluchowski, 2006). Therefore, a DWH is a “subject oriented, integrated, non-volatile, and time variant collection of data in support of management decisions” (Inmon, 1996, p. 31). Therefore, the most important task of a DWH is to structure and harmonize data from different operational systems. It is an integrated collection of data extracted from operational, historical, and external data bases; this data is then cleaned, transformed, and catalogued for retrieval and analysis (Simonovich, 2006) in order to provide BI for business decision-making.
Figure 2.3 puts the concepts of DWH and analytical processing into context. As the figure illustrates, a DWH stores data that has been extracted from various operational and external data bases of an organization and can include: AOKN’s doctor data base, hospital or rehabilitation information, or external individual programmed data base for specific health care programs of an organization. It serves as a central source of the data that has been cleaned so they can be used by managers and other business professionals for analysis purposes.
Figure 2.3: DWH structure (according to O’Brien, 2004)
As depicted in Figure 2.3, there are different ways for managing the contents of a DWH. Apart from ad-hoc access to a DWH, the use of analysis tools is popular. A multidimensional structure can be a data base model that uses multiple dimensions to represent data. Such a structure can be appropriate way for end users to develop individual analysis on a multidimensional model. These aggregated data structures can be seen as data cubes appropriate for BI end users. Online analytical processing draws on four commands to navigate through analysis:
slice: extracting a view by omitting the data of irrelevant dimensions
dice: turning the cube to change the two dimensions desired to be seen on screen. It is not possible to visualise three or more dimensions.
drill down: looking into details, which corresponds to looking into a cube within a cube
roll up: aggregating data
In contrast, a relational data base is a logical data structure in which all data elements within the data base are stored in the form of tables. Most commercial data bases today are relational structures or are based on these. In the relational model, all data elements within the data base are interpreted as a set of simple
tables. In information systems design and theory, as initiated at the enterprise level, the “Single Point of Truth” refers to the practice of structuring information models and associated schemata such that every data element is stored only once. That means that the element is positioned in no more than a single row of a single table. Organizations are looking for and need BI solutions that can deliver information based on a single view of the operational data, rather than multiple views and resultant inconsistent information (Logica BI Framework, 2010). The different navigation opportunities of different data models - aggregated cubes or relational in-depth data files - to drill down and up will be a crucial point for BI end users that have to be enhanced later.
On the other hand, data that resides outside of structured data bases or DWH is called unstructured data. This includes electronic documents, Powerpoint presentations, email, images, schedules or multimedia files. This data typically resides on individual computers or on file servers. In some cases, when the unstructured data is particularly important to the company and it needs to be searchable or requires further analysis, it might be organized into a structured data base and made available as part of a BI solution. There are a number of so-called content management systems that are designed to organize unstructured data in order to help control and manage content, versioning, and access rights (Brannon, 2010). However, these data structures can also be integrated into a DWH without implementing an additional content management system.
In summary, a DWH is a structured data base for collecting and revising relevant data from various sources. This instrument enables companies to further integrate data forms from more external sources to create benchmarks relative to competitors and/ or care providers. In this way, data structures can be harmonized for easier and better performance access and analysis of different constellations. It will also be necessary to implement standardized data dimensions and ratios and to document them appropriately for other users. With regard to target-oriented and data-oriented content, the DWH is determined through the core elements topic orientation, consolidation, constancy and time orientation (Gansor, Totok & Stock, 2010). The DWH is a basic requirement for a structured data generation from many operational sources with many different data structures and formats. In order
to avoid protracted data processing, the DWH functions as single point of truth and basic data orientation for BI end users. These infrastructural requirements constitute the BI framework for implementing further business and BI strategic objectives. Thus, a DWH is an important basis for the conceptual framework, which will be developed in Section 5.2.
Good architectures address the cost, benefits, and risks of every design decision. Good architectures draw upon existing skills and tools where they make sense and add new ones where needed (Lopez and D’Antoni, 2014). For most business users, the DWH’s front room with its BI reports and analytic applications is the only visible layer of the DWH. It is possible to build BI applications without the benefit of a DWH, but this rarely happens. A well-built DWH adds value through the dimensional model and load processes, thus it makes no sense to replicate this effort to build a standalone BI application. Most successful BI applications are an integral part of the user-facing portion of the DWH (Kimball et al., 2010). These BI technologies and tools will be discussed in Section 2.1.3. Serving big data, with BI technologies and tools, and the data provision will also be an important topic that will be presented in the next Section.