​​Predicting Passenger Willingness to Pay for Ancillary Services on the BCN-LGW Route - a Machine Learning Approach Using Likert-Scale Survey Data​

Abstract

​​The growth of ancillary services has become a central driver of airline profitability, yet understanding passengers’ willingness to pay (WTP) for these offerings remains a challenge. This thesis investigates the key factors shaping WTP for ancillary services among travellers on the high-traffic Barcelona (BCN)–London Gatwick (LGW) short-haul route and evaluates the potential of predictive modelling to forecast individual passenger WTP. Deploying a structured survey (N=171), the study collects and analyses data on passenger demographics, travel behaviour, and attitudinal responses toward four key ancillaries: seat selection, checked baggage, in-flight meals, and Wi-Fi. A preprocessing pipeline encodes Likert-scale responses and constructs composite WTP scores, which serve as targets for a range of machine learning algorithms (Logistic Regression, Random Forest, Support Vector Machine, and XGBoost). Model performance is evaluated using metrics such as accuracy, F1 Score, and ROC-AUC, with further interpretability gained through SHAP values and permutation importance analysis. The results reveal pronounced heterogeneity in WTP: seat selection and checked baggage appeal broadly, while in-flight meals show lower and dispersed WTP, and Wi-Fi has sharply polarised demand. Statistical and modelling analysis show that travel purpose, annual income, and age are the most significant predictors of ancillary WTP, while gender has negligible influence. Predictive models achieve strong performance, especially for passengers with clearly high or low WTP; however, the ’medium’ WTP segment remains challenging to classify, suggesting opportunities for richer data and model refinement. These findings highlight the potential of dynamic, segment-driven ancillary pricing and marketing. The study calls for strategic product design, targeted offers, and investment in predictive analytics, allowing airlines to align their ancillary strategies with evolving passenger needs. Limitations include sampling constraints and reliance on self-reported data; and future research should incorporate revealed-preference data, longitudinal analysis, and broader feature sets to further advance ancillary revenue management.​