As an airline that traditionally invests in science and research for its revenue management practice, Etihad Airways chose to work with PROS to identify, test and improve key features of the industry-leading PROS Bayesian forecaster as part of its ongoing partnership. This process involved almost two years of collaboration to identify improvements, implement new ideas into the product, and deploy and test the impact of these changes.
The Bayesian forecaster has been a core component and differentiator for PROS Revenue Management. It has driven proven incremental revenue for more than 70 airlines around the world. In collaboration with Etihad Airways, PROS sought to implement new features such as the Customer Choice Model, Detailed Closure Probability, Unconstraining Attenuation, and Data Collection Point (DCP) Segmentation and gather valuable feedback from an airline that operates in a complex, dynamic environment.
CHALLENGES
Etihad Airways’ solutions and innovations team within its revenue management department was exploring opportunities to increase the revenue management forecaster’s accuracy. It examined several areas of its network to identify specific improvements. For example, lower, highly price-sensitive classes had high constraining. Etihad Airways saw an opportunity to further explore unconstraining attenuation and closure probabilities to address this issue. Another area of improvement included DCP Segmentation to address where the impact of observations across long range of days prior was too large. Finally, there was an opportunity to address situations where demand on connecting itineraries did not always choose the shortest travel time. Etihad looked to PROS to help identify the best solutions for these areas of research.
SOLUTION
At the beginning of this project, PROS and Etihad Airways used real data to test specific improvements to the forecasters in a lab setting. Together, they identified improvements that provided benefits to forecasting performance and then wrote requirements and implemented improvements that they believed would show value.
Based on these early tests, the team saw significant improvements to forecasting performance that warranted implementation of the features into the PROS Revenue Management product. Etihad Airways and PROS agreed to implement the following features:
- Customer Choice Model: Using a history-based model to understand the choices passengers make between itineraries to disaggregate forecasts to the lowest level.
- Percentage of Time a Class is Available in Unconstraining: Using more accurate availability information in the unconstraining process, which results in more accurate unconstrained forecasts.
- DCP Segmentation: Splitting DCPs into groups (or segments) of periods with similar behavior, allowing for more appropriate sharing of information across DCPs.
- Unconstrained Attenuation: Enabling more control and flexibility on the level of unconstraining that is calculated on a market/class level. In some cases, the unconstraining captured by the system may be too high in the lower classes, so airlines can scientifically dampen the unconstraining with this tool.
In addition to implementing these features, Etihad Airways underwent a detailed study with the PROS science team, which provided new values for seasonality settings to better capture seasonality, new DCPs to better segment passenger arrivals, and new days prior patterns in DCPs to better capture demand at each DCP.
RESULTS
The PROS and Etihad Airways teams collaborated to implement these new features in the Etihad environment and tested for nearly seven months to take advantage of the combined feature effects and report the benefits. For the testing, Etihad Airways benchmarked the new forecasts against the old forecast as well as against expected results.
Highlights of the results included the following:
- System forecast showed major improvement by reducing double counting in forecasts on tactical classes.
- Seasonality factors showed a major improvement after the new science study, thus enabling Etihad to capture peaks and valleys more accurately.
- Several markets have shown major improvements in unconstrained demand levels, forecast volumes and seasonality across departure year, which almost warrants nil/minimum user intervention.
CONCLUSION
Science, the foundation of PROS Revenue Management solutions, was a key driver in the successful implementation of new features and resulted in direct business value for Etihad Airways. The close collaboration between the airline and PROS yielded best practices and improvements to PROS Revenue Management that can be used and modeled by airlines around the world.