The effects of different levels of normalization on data entry and data entry controls : an experimental evaluation

End users are using relational database software resident on their personal workstations to create their own tables and update an increasing volume of data. The value of these databases depends on the accuracy of the data they contain. Unfortunately, end users may not have the training or expertise to normalize the data structures or implement input controls so that data accuracy is maximized. This thesis reports the results of an experiment that examined the effects of normalization level and input controls on data entry errors. The experiment also examined the effects of information systems background on the ability of end users to specify data structures and input controls. To test the five hypotheses proposed in the thesis, eighty participants were randomly assigned to one of three data structures: not normalized, first normal form, and third normal form. The first hypothesis proposed an inverse relationship between normalization level and the proportion of errors users enter into a database. Hypothesis two anticipated a direct relationship between normalization level and data entry speed. The third hypothesis stated that data validation procedures reduce the proportion of errors users enter into a database. Hypothesis four proposed that users with more extensive information systems backgrounds would identify more input controls. The fifth hypothesis stated that users would request more data structure changes for less normalized databases. The results of the experiment provided support for all five hypotheses. Although normalization improves data accuracy, well designed input controls appear to provide as great if not greater improvements to data accuracy. Even the combination of high levels of normalization and stringent input controls. however, would not prevent all data entry errors. One implication of the results is that assistance and training in creating normalized data structures and designing input controls should be provided to end users who are creating and maintaining their own tables. Another implication is that additional quality control procedures are needed to ensure data accuracy.