The objective of this study is to explore the impact of the structure of a Bayesian Network on its overall run-time, potential unwanted bias, and accuracy in performing a classification task. A credit card default dataset was utilised to construct six networks with varying structures and to learn their conditional probability distributions. The resulting networks were analysed, potential discrepancies were identified, and experiments were conducted to quantify any differences. The findings suggest that the structure of a network can significantly impact its performance, and further studies should focus on validating the statistical significance of the results.