Wizz Air is the largest Central and Eastern European low-cost airline with over 500 routes from 25 regional bases. As a value-driven airline, it focuses on innovation all along the customer journey. Its goal is to provide affordable services with great user experience for its passengers.
To successfully represent these core values, such as innovation and optimized user experience, it is critical for the company to understand its customers’ needs by analyzing relevant data and providing personalized interactions based on the insights gained. One of the customized experiences that Wizz Air offers is the personalized recommendation of ancillary products, such as seats and priority boarding.
Wizz Air was carrying out non-personalized email campaigns prior to departure. The performance of these campaigns was below expectations and Wizz Air was looking for a way to improve the email campaigns with personalized messages and recommendations.
Wizz Air was also using the Google Analytics 360 Suite to record user interactions on its online platforms but they had not yet made use of the stored data.
The challenge, thus, was fourfold:
- Extract useful features from Google Analytics 360.
- Segment customers to obtain targetable groups.
- Recommend relevant ancillaries to relevant customers.
- Demonstrate the improvement of personalized messages ensuring the performance was AB tested.
Since Google Analytics 360 data already resided in Google BigQuery, the ETL pipeline for obtaining useful features was also designed and implemented in BigQuery. As Google Analytics 360 data wasn’t structured to have the features directly, it took several transformation steps to extract context data, historical user behaviour (purchase and search), and historical flight data.
Given that some of these steps are interdependent, we used Google Cloud Composer to orchestrate the tasks. Cloud Composer allows the user to easily monitor the execution and alert them in case of errors.
In general, it is advisable for a user segmentation to be interpretable by the business. By getting the extracted features right, the customers could be divided into segments that are both targetable and meaningful. These two aspects were fundamental for message copywriting.
It was also essential to recommend relevant items to the customers. Since there were not many items to recommend, we devised a separate propensity scoring algorithm for each ancillary, which allowed more flexibility in the modelling; different features proved to be of use for different ancillaries. The scoring models were trained on AI Platform.
We also used Cloud Composer to execute both the segment assignment and propensity scoring tasks.
By implementing a personalized email recommendation pipeline, Aliz has helped Wizz Air in taking the first steps toward the goal of reaching its customers with relevant messages. The post-booking ancillary conversion rate increased by 21% showing the enormous potential of personalized recommendations.
Aliz contributed to setting Wizz Air off on acquiring a data-driven approach in serving its customers and provided hands-on experience in implementing it.