FLYR for Hospitality Now Available on Oracle Cloud Marketplace Arrow

Resource Hub / Airlines / Revenue Optimization / Low-Risk Revenue Management (RM) Optimization

Low-Risk Revenue Management (RM) Optimization

Low-Risk Revenue Management (RM) Optimization: Evaluating Model Performance Prior to Implementation

Antiquated revenue management software, relying on legacy forecasts, has limited airlines’ ability to make critical commercial decisions. This has resulted in significant gaps in revenue and business growth opportunities, particularly in the face of today’s volatile travel environment.

Adopting a fully cloud-based SaaS solution allows for a more nimble and strategic data management experience for airlines – from the first stages of implementation to regular use of the tool. Additionally, with a more data-efficient approach and access to more contextual data, advanced Revenue Management (RM) deep learning models have been shown to improve forecast accuracy and provide additional revenue uplift.

Attempting to deploy changes to an existing Revenue Management System (RMS) or implementing brand new RM strategies can put significant revenue at risk. So how can you measure the revenue potential of a new RM strategy prior to its implementation? One way in which FLYR achieves this is by utilizing Reinforcement Learning (RL) and Offline Policy Evaluation (OPE) techniques to estimate the revenue performance of any model prior to A/B testing. As a part of our rigorous model validation process, this approach allows us to test only the most promising models and iterate faster, all while risking less revenue.

Reinforcement Learning (RL)
Reinforcement Learning (RL) is a subfield of machine learning and is concerned with optimizing a series of decisions. It does this typically by estimating the value of each possible decision in the context of some behavior policy or strategy. These techniques are often used to find the optimal behavior policy (the optimal pricing strategy in the case of airline RM), but they can similarly be used to estimate the value of existing strategies. When applied to airline RM, we can learn from historical data to estimate how much a novel pricing strategy might be worth.

Offline Policy Evaluation
Once a new RM deep learning model is developed and ready for deployment, we use OPE techniques to judge if it is worth more revenue than the existing model. If it is estimated to be more valuable, then we move forward to an A/B test and implementation. And if not, we go back to the drawing board and attempt to train an improved model. This ensures that we do not waste valuable time and revenue testing with subpar RM models.

A science-first approach is the foundation to a more effective revenue management system. With the right technology, data, and modeling approach, airlines can implement new RM strategies with confidence and unlock significant revenue potential and influence decisions across all commercial functions.

The FLYR Difference
FLYR’s Cirrus Revenue Operating System includes four critical components: a single repository and unified data model to integrate all airline data, ultra-accurate revenue and load factor forecasts utilizing advanced deep learning AI technology, continuous class-agnostic pricing algorithms, and end-to-end reporting, analytics, and controls for airline decision-makers. Contact us for more information including details about FLYR’s 12-week no-risk implementation process.

Similar stories

Early adopters of AI across industries have already realized and demonstrated its significant benefits, but the travel industry is still slow to embrace this cutting-edge technology to improve business operations, customer experiences, and revenue performance.
Why this legacy airline is transitioning to a low-cost model, and how it will continue to be agile in a changing industry.
While the aviation industry as a whole has been cautious to make the switch from legacy tools and techniques, digital-first airlines have unlocked incredible value across commercial functions by leveraging the latest in AI.