This paper introduces a new method for estimating sell-up rates that show a passenger’s propensity to purchase the upper fare class using class-based historical booking data. The performance of models in the literature, such as Direct Observation (DO) and Inverse Cumulative (IC), is not satisfactory when applied to the actual historical booking data since their results do not match the business expectations. We improve these models and design data pre-processing and solution techniques that can be beneficial for practitioners in dealing with historical booking data. Adjusting the past bookings with fare class availability data, our model provides a robust sell-up rate estimation. We examine data of a major European airline in our analysis. Numerical results show that our proposed method decreases MAPE (mean absolute percentage error) significantly.