Narrative

#1) High Correlation Between Departure and Arrival Delays, Strong Link Between Online Booking and Boarding, and Removal of Low-Impact Features Like Age and Flight Distance

The Correlation Heatmap provides a comprehensive overview of how various operational metrics and service quality variables relate to one another. The most striking correlation appears between Departure Delay and Arrival Delay, with a coefficient of 0.96. This indicates that these two variables are almost interchangeable, which aligns with the operational reality that delayed departures typically lead to late arrivals. It also reinforces past findings by Liu & Zeng (2007) and Li et al. (2015) that operational disruptions are tightly linked and have cascading effects on customer satisfaction.

Another critical cluster emerges in Ease of Online Booking and Online Boarding, exhibiting a strong positive correlation of 0.71, suggesting that passengers who find the ticket purchase process seamless also tend to rate the boarding experience positively. This mirrors Ban and Kim’s (2019) emphasis on the importance of seamless digital interactions and confirms that user experience in pre-boarding stages plays a crucial role in shaping overall satisfaction. This implies that improvements in the digital airline’s app or website could simultaneously enhance multiple customer experiences.

In contrast, features like Age and Flight Distance show very weak correlations with the rest of the dataset, meaning they operate independently and are unlikely to distort model behavior through multicollinearity.

Overall, the heatmap helped us identify which variables could be removed, combined, or retained. It served as an essential tool in shaping our preprocessing decisions and allowed us to focus the model on features that carry distinct, non-redundant information.

#2a) Higher Comfort Comes with a Better Satisfaction

The Class by Satisfaction chart shows a strong positive correlation between cabin class and passenger satisfaction. Business class passengers reported the highest satisfaction rates, with more than twice as many passengers satisfied compared to those who were neutral or dissatisfied. In contrast, the majority of Economy and Economy Plus passengers expressed dissatisfaction, while only about one in five travelers reported a positive experience. These findings support Ringle et al. (2011), who emphasize the importance of class of service and travel purpose in shaping customer satisfaction, particularly for leisure travelers, who are especially sensitive to perceived comfort and safety.

This stark contrast suggests that cabin class is not only a proxy for comfort but also for service quality and perceived value. It reinforces the idea that passengers who pay a premium expect, and likely receive, more responsive in-flight attention, fewer disruptions, and better overall treatment. Those who can afford business class are not just paying for comfort; they are buying access to a social environment where their needs are prioritized. This reinforces a broader economic disparity, where higher satisfaction is effectively purchased, while those with fewer financial resources are more likely to encounter crowding, delays, or inattentive service.

The near-identical dissatisfaction rates between Economy and Economy Plus also suggest that mid-tier options may fail to meet expectations. Passengers who pay more for an “upgraded” experience may still feel underserved if the improvements are marginal or not clearly communicated. This gap between expectation and reality may increase frustration and contribute to lower satisfaction, especially for those seeking affordable upgrades as a middle ground between price and comfort.

#2b) Travel Purpose Drives Satisfaction: Business Over Leisure

The Type of Travel by Satisfaction chart suggests that travel purpose significantly shapes how customers evaluate their flying experience. Business travelers are more likely to report satisfaction compared to leisure travelers, even though they likely face similar flight conditions. One possible explanation is that business travelers often enjoy additional perks sponsored by their employer, such as premium seating, priority services, and streamlined check-in processes.

Personal travel, on the other hand, is often funded by the traveler themselves and carries emotional expectations, such as relaxation, family reunions, or vacations. When delays, discomfort, or poor service occur, the disruption can feel more personal and disappointing. Leisure travelers are also less likely to receive premium treatment or loyalty benefits, which may contribute to their lower satisfaction. Bezerra et al. (2021) further highlight that non-frequent travelers hold different service expectations, making them more sensitive to unmet promises or discomfort. Leisure passengers also tend to receive fewer loyalty perks and less personalized service, reinforcing the satisfaction gap.

This disparity raises important considerations for equity and customer service in the airline industry. If airlines are more focused on serving business travelers because of their higher spending, personal travelers may be left with a lower quality of service. The data highlights the need for airlines to reassess how they serve non-business passengers, particularly during peak travel seasons when stress levels are higher and service inconsistencies are more likely. Addressing this imbalance could help ensure a more equitable experience for all types of travelers.

#2c) Frequent Flyers Enjoy Higher Satisfaction; New Travelers Left Behind

The Customer Type by Satisfaction chart emphasizes the critical role of customer loyalty. Returning customers showed much higher satisfaction than first-time flyers. Returning customers are far more likely to report a positive experience, while first-time flyers overwhelmingly report dissatisfaction. This suggests that repeat customers benefit not only from familiarity but also from loyalty programs that reward them with upgrades, faster service, and preferential treatment.

This finding echoes Leon & Martín (2020), who demonstrate that satisfaction often stems from a mix of technical factors (like efficiency and punctuality) and functional ones (like recognition and personalized service), both of which are more available to frequent flyers. Airlines are incentivized to retain loyal customers by offering perks such as lounge access, faster boarding, and personalized offers. However, Sezen et al. (2018) emphasize that these loyalty-driven benefits can skew service quality toward certain demographics or travel patterns, leaving newer or infrequent travelers with less support and lower perceived value.

This issue reflects how trust and loyalty are built into the airline system. Airlines are more likely to invest in keeping frequent flyers happy, offering perks like lounge access, early boarding, and seat upgrades. First-time travelers, on the other hand, may be unfamiliar with how to navigate airport logistics or how to access better service. They are also less likely to receive priority treatment, which could make them feel overlooked or under-supported.

From a broader perspective, this points to a service system that favors existing relationships and potentially disadvantages new or infrequent travelers. It suggests that customer experience is not just shaped by the flight itself, but by a traveler’s accumulated history with the airline. While loyalty programs are valuable marketing tools, airlines must be mindful not to let them reinforce inequality in the passenger experience. Improving support for first-time passengers could not only boost their satisfaction but also help build long-term loyalty and trust.

#3) High Accuracy, High Impact: Using XGBoost to Predict Customer Satisfaction

We compared four classifiers—Decision Tree, Random Forest, AdaBoost, and XGBoost—on our dataset. XGBoost emerged as the top performer, achieving an overall accuracy of 96.28%. Its confusion matrix breakdown shows that out of all negative instances, it correctly labeled 14,429 as negative (true negatives), and out of all positive instances, it correctly labeled 10,591 as positive (true positives), demonstrating both high specificity and sensitivity.

This finding underscores the growing role of advanced ensemble models in customer experience research. As Syed et al. (2024) demonstrate using deep learning methods like BERT and PEGASUS, customer satisfaction can be reliably predicted using review data. Similarly, Srinivas & Ramachandiran (2020) show that machine learning applied to online airline reviews can uncover systemic service issues and identify priority areas for improvement.

This result supports the value of using advanced ensemble methods for prediction in customer experience research. The model’s strong performance not only confirms that passenger satisfaction is predictable from service and operational variables, but also suggests that machine learning can help airlines identify improvement areas before dissatisfaction becomes widespread. With further tuning or real-time deployment, models like this could help prioritize service recovery efforts or proactively target at-risk customer segments.

#4) Digital First: Wi-Fi and Online Boarding Surpass Traditional Amenities in Driving Satisfaction

The feature chart highlights which variables most strongly influence the model’s prediction of passenger satisfaction. In-flight Wi-Fi Service stands out as the most impactful factor by a large margin, contributing nearly half of the total predictive power. This suggests that digital connectivity has become a core expectation for modern travelers, outweighing even traditional service elements like seat comfort or food. Following Wi-Fi, Online Boarding, Type of Travel, and Class also show meaningful influence, reinforcing that both digital experience and travel context shape how passengers perceive their flight. This trend is strongly supported by Park et al. (2020), who found that while traditional service aspects like punctuality and staff courtesy still matter, satisfaction is increasingly driven by how smoothly the full journey, especially digital touchpoints, is managed.

In contrast, variables such as Check-in Service, Leg Room, Cleanliness, and Entertainment have minimal importance in the model. These insights suggest that airlines aiming to improve satisfaction should prioritize investments in Wi-Fi reliability and user-friendly digital interfaces, especially for economy and leisure travelers who report lower satisfaction overall.

These results suggest that airlines should rethink their investment strategies, particularly for economy and leisure travelers, who tend to report lower satisfaction and are less likely to receive high-touch service. As Ban & Kim (2019) note, passengers increasingly mention digital convenience in online reviews, meaning that improving Wi-Fi reliability, app interfaces, and self-service tools may deliver stronger satisfaction gains than marginal upgrades to physical amenities.

Finally, the feature heat map also helps identify redundancies and multicollinearity. It offers critical insight into where airlines can focus to enhance both perception and performance.