Design a system to predict YouTube ad conversions.

Focus on outlining the architecture of a predictive model that can analyze various data inputs related to YouTube ads, such as viewer demographics, ad engagement metrics, and historical conversion rates. Discuss the types of algorithms you would consider for prediction, such as regression models or machine learning techniques, and justify your choices based on the nature of the data. Address how you would gather and preprocess data, including handling missing values and feature engineering. Finally, consider how you would evaluate the model's performance and iterate on it based on feedback and new data.