Organizational investment in data repositories constitutes a massive share of organizations overall investments (Cooper et al., 2000; Fisher et al., 2003). Decision makers use data retrieval tools to generate information from data repositories. The quality of the information retrieved contributes significantly to the organizations success or failure. Decision makers use this information in analytical tools to derive significant business trends and opportunities.
Poor information quality can lead to costly decisions and, hence, to reduced profitability. Inaccurate information, however, is not necessarily a database-specific related problem. Data repositories may exhibit near perfect data quality, but users may still formulate wrong queries and thus retrieve inappropriate information.
Prior research examining end users' data retrieval performance has shown that there is a high incidence of errors during query formulation. Prior research investigated users' overall query performance using parsimonious data models (PDMs) and ontologically clearer data models (OCDMs). Poor data retrieval performance by end users can be attributed to the insufficient comprehension of the data models that represent the information or their poor technical SQL skills. End users respond to information requests in two stages. First, they must jointly map the information request and the data model. Second, they formulate the SQL query to retrieve the information needed.
This study focuses on stage one: jointly mapping the information request and the data model. It examines the impact of larger data model representations, PDM vs. OCDM, on end users' performance when undertaking component, record, and aggregate level tasks. The Bunge-Wand-Weber's (BWW) ontology was used to formulate nine hypotheses. Experimental results indicate that, on average, users achieved higher percentage scores and took less time to analyze each information request when using the ontologically clearer data structure. In addition, the results indicate that participants using PDMs significantly outperform participants using OCDMs for component level tasks when optional relationships in the PDM were converted into subtypes in the OCDM. Participants using OCDMs, however, significantly outperform participants using PDMs for record and aggregate level tasks when optional relationships were converted into subtypes.
This research can aid practitioners design data structures and select query tools that optimize end users' query performance. For researchers, the study extends prior research in end users' query performance by investigating the mapping between information requests and data model components during the query composition process for larger application domains.