A framework for the selection of appropriate data modelling approaches in data warehouses

Abstract

Determining which data modelling approach will suit a company’s data, information and business intelligence needs best can be a complex process. Many companies today have not yet adopted the correct data modelling approach, which has a direct impact on the decision making in the company. A data warehouse is an essential component of business intelligence. The need for information contributing to decision making within a company is the only constant in an ever-changing world of business. The main research objective of this study is the creation of a framework that guides the selection of a suitable and feasible data model for a data warehouse that has a direct impact on the business intelligence system. In this study, the research paradigm classification scheme determined by Van der Merwe et al. (2020:2) is followed and refers to four research paradigms, namely positivism, interpretivism, critical theory and design science research. For this study, design science research was adopted as a research paradigm, as it involves a rigorous process to design artefacts that are intended to solve observed problems. Design science research is also becoming a popular research paradigm in information systems research. A research plan was developed to assist the researcher during the study. The research plan involved three phases and was based on a design science research framework. By adopting design science research, the researcher can evaluate the designs and communicate the results to appropriate audiences as well. A survey was conducted to identify which data warehouse and business intelligence system elements would be challenging and to gain much-needed knowledge on issues, such as additional factors that need to be considered regarding the data warehouse or business intelligence system. To complement the survey research method online questionnaires was chosen as data generation method. The total number of questionnaires returned from employees currently employed in the information technology industry was 112. Data sets of three different companies were used and data models were designed and developed for each company’s data sets. The companies willing to participate and share their data were referred to the researcher by participants of the questionnaire. By identifying the best-suited data model for a company, a data warehouse was designed and possible requirements set by the company were addressed. A second questionnaire, sent to representatives of the three companies, was used to determine the users' experience of the developed data models and to determine if there were any further suggestions for improvement. A total of 68 respondents completed the second questionnaire. The second questionnaire was used as a quality measure to test rigor and robustness of the designed artefact. When choosing a data modelling approach, a company should, along with other factors, also take into consideration the impact that the data model has on business intelligence. Data modelling is an ongoing process that should be revisited when changes are required in the data warehouse and business intelligence system. A data model should be re-evaluated when changes occur either in the external environment or internal environment of the company, for example, in the company’s business information needs, data parameters, enterprise size, business processes or current data architecture. Companies can use the developed framework, which consists of 15 guidelines, as a tool or guideline when selecting a data modelling approach for their data warehouse. The framework was developed based on initial guidelines identified throughout the study. Three themes were identified within the initial guidelines: project management and communication, data modelling approaches and data model qualities, and lastly, data warehouse and business intelligence elements. This study contributes to knowledge regarding data modelling for organisational use and provides organisations with a framework and process which can be used to determine which data modelling approach is most suitable for their business.