console.warn && console.warn("[GTM4WP] Google Tag Manager container code placement set to OFF !!!"); console.warn && console.warn("[GTM4WP] Data layer codes are active but GTM container must be loaded using custom coding !!!");

Our projects

AI to improve demand forecasting in the airline industry

Improve the accuracy of demand forecasts on complex air routes, to improve revenue management and maximize operational efficiency.

Development of an AI solution that easily integrates into managers' day-to-day operations, providing accurate recommendations without overloading their workflow, while improving demand forecasting.

The significant improvement in demand forecasts has led to a measurable increase in revenue, supported by a precise methodology, demonstrating the effectiveness of the solution on hundreds of routes.

corner-image corner-image
Takeoff – Developing a tool to improve demand forecasting

Most airlines use a complex revenue management system (RMS) that analyzes data like aircraft capacity, operational costs, schedules, and competitors’ offerings. These systems also estimate future demand. The RMS analyzes this information to identify the optimal price for each seat on every flight at any time, thus optimizing aircraft occupancy.

 

However, the demand forecasting module of this airline’s RMS did not meet expectations during periods with significant variations in passenger volume.

 

To improve their forecasts, IVADO Labs developed an artificial intelligence solution. This recurrent neural network was trained on a dataset containing historical booking data and temporal information to contextualize the bookings.

The AI solution was deployed as a pilot project on several different routes, where it improved predictions by an average of 25% compared to the airline’s RMS module. In some cases, the improvement was as high as 40%.
Cruising speed – Integrating AI intelligently

IVADO Labs designed the AI application in collaboration with the airline’s demand managers.

 

These analysts manage a vast portfolio of routes, each with unique characteristics and market dynamics, and due to time constraints they often have to focus on a subset of routes. The goal was to enhance their capabilities and simplify workflows without overwhelming them with data.

IVADO Labs developed a solution to integrate the new forecasts into the managers’ daily operations and apply various adjustments (also called influences) to the RMS to align its forecasts with those of the AI tool.

 

This “influence optimizer” presents information to analysts via an intuitive and interactive visual interface, highlighting areas where adjustments would have the greatest impact.

Demand managers retain the final say, however, with the ability to apply recommendations or modify them based on their knowledge and experience.

 

For the airline, this seamless approach facilitated change management while mitigating the risks associated with adopting a completely new RMS.

Turbulence: Predicting in the uncertainty of COVID-19

By March 2020, the project was well into the testing and evaluation phase, with initial results indicating a significant improvement in forecast quality.

 

However, the COVID-19 pandemic created an unprecedented situation in civil aviation, with a drastic drop in demand and historical booking trends instantly becoming obsolete. As it was trained on pre-pandemic data, the AI solution developed for the client was initially unable to adapt to this new reality.

 

IVADO Labs and the client had to adapt quickly. They implemented a data augmentation strategy, leveraging studies by the International Air Transport Association (IATA) that predicted the pace of global air traffic recovery.

This technique involved generating synthetic data that would reflect expected demand during different phases of the pandemic: the initial drop in passengers, the stabilization plateau, and the gradual recovery.

Fictitious flights and bookings representing these scenarios were integrated into the model’s training dataset, enabling it to learn how to respond to extreme events and better predict demand in uncertain contexts.

Integrating synthetic data into the AI model’s training dataset proved to be an effective strategy for improving its ability to predict demand in atypical situations, without compromising its accuracy on pre-pandemic data.

 

Robustness and resilience are crucial characteristics of the solution developed by IVADO Labs. Other “black swan” events could occur in the future, and even without such events, the aerospace industry is not immune to turbulence. According to Airports Council International, international air traffic is expected to grow by 5% annually until 2042.

 

corner-image corner-image
Landing: Measuring the impact of the AI solution on revenues

The AI solution developed for the airline improves its demand forecasting, but does that translate into increased revenues?

 

To quantify the financial impact, IVADO Labs uses a deep-learning-based methodology to estimate counterfactual revenue, or the revenue that would have been generated without the AI solution’s intervention.

We have developed an algorithm to assess the precise impact of our solution by comparing actual revenues with counterfactual forecasts, i.e. with what revenues would have been without our intervention. This precision is a valuable asset in the field of revenue management, where every percentage increase can have a significant impact on the bottom line.

Initially reserved for about sixty routes, the AI solution now predicts demand for hundreds.

 

With the knowledge transfer and support we provided, the airline was able to manage a portion of the deployment internally. This embodies our mission: to harness AI to support companies’ ambitions and then help them become autonomous in using these new tools.

Beyond the demand forecasting tool, the collaboration between IVADO Labs and the client also paved the way for other projects aimed at maximizing the company’s revenues, including an AI-powered solution to optimize seat pricing.