Click-Through Rate (CTR) prediction is an important application in online advertising, and deep learning-based models are developed to maximize CTR prediction. In this study, the effect of normalization methods and different placements of the layers implementing these methods on CTR prediction performance are investigated. Batch Normalization and Layer Normalization, which are often employed in deep learning- based methodologies, were used as normalization techniques. The experiments are conducted on Avazu and Criteo datasets, which are widely used in CTR prediction, as well as a company dataset. Experiments on different models developed for CTR prediction have shown that the correct normalization method and its use in the right place can show a relative improvement of up to 0.6583%, demonstrating the importance of using normalization techniques.