Revenue management systems are hardly new, but in today’s fluctuating market, existing RM systems no longer deliver the optimum results needed by airlines today. At FLYR Labs, we are working to improve upon traditional revenue management (RM) systems, which are outdated and fail to utilize the new technology required to create hyper-accurate forecasts and revenue optimal pricing strategies.

The Revenue Operating System^® from FLYR overcomes legacy systems’ limitations by making AI, and more specifically deep learning, central to revenue optimization. By harnessing the power of deep learning, artificial neural networks, and cloud computing, The Revenue Operating System overcomes the inevitable errors that come with legacy RM systems giving airlines a winner-takes-all advantage.

How does FLYR harness deep learning? What makes The Revenue Operating System unique?

Traditional RM systems rely on a multi-step model, a method of forecasting that was ‘state of the art’ three decades ago but has failed to keep up with changes in the travel industry and society. These systems were designed to make future predictions via sequential modeling, in which each piece of contextual information produces an output that informs a secondary decision. There are several limitations to this method. First, the if-then outcomes resulting from this method are too linear, only taking into account surface historical data to provide a price. Second, errors made early on perpetuate throughout the whole process, making it impossible to know if forecasts are correct. Quality becomes increasingly hard to measure, and introducing new contexts is almost impossible.

FLYR’s use of deep learning sets us apart from these outdated legacy systems. This cutting-edge technology unlocks the full potential of data, thereby revolutionizing revenue optimization for airlines and other organizations. Our system learns from input context to continuously improve operations, without the need for human interference. Processing larger amounts of rich data provide better predictions, resulting in optimal decision making and allowing analysts to focus more on outlier outcomes and less on standard pricing. More profitable decisions attract more users, which in turn generate more data, creating a virtuous circle of revenue optimal pricing and hyper-accurate forecasting.

Another important factor differentiating FLYR’s system is the integration of ‘reward engineering.’ This makes it possible to train the network to identify events requiring context-specific output, whether to maximize revenue, improve conversions, or reduce costs.

For example, each FLYR airline customer has a breadth of historic flight data which is input into our system. Each flight creates one ‘observation’ that can be tracked. From this data point, The Revenue Operating System considers the context for the flight, with 56 signals feeding into this process. We understand the importance of context when it comes to future decision-making, so our operating system recognizes each flight’s landscape based on past events, competitor data, similar flight times and origins and destinations, and more, and learns how changing contexts will affect the final decision.

In addition to seeing context, our technology looks at which strategy was employed and whether a decision had a positive or negative impact on overall revenue. The system will assess over a billion decisions, training itself to recognize the patterns of reward in real time. Again, accuracy increases as more data becomes available, leading to unlimited levels of actionable insight.

Deep learning and neural networks

To understand The Revenue Operating System, think of it like a human brain. Multiple functions and neural networks collaborate to produce multiple outcomes. One part of the brain, known as the dynamic neural network, learns from multiple pieces of information, such as competitor pricing and booking arrivals, and how these impact one another. The second part of the brain, known as the static neural network, uses a geographical encoder and calendar encoder to learn how the market changes over time. At the top level, the network integrates all the information to produce long-short-term-memory (LSTM) that predicts the best next move, including fare pricing and flight forecasting.

It is important to note that the Revenue Operating System is not designed to replace the human analyst. It is a strategic decision support tool that enables revenue managers to maximize their effectiveness. AI eases the revenue analyst’s workload, managing 90 to 95 percent of the flight and enabling the analyst to focus their expertise on the value-adding remainder by trying different strategies for each scenario. The system’s deep learning function then assesses whether the analyst’s chosen strategy was effective and should be replicated by the system in the future.

Supported by our technology, analysts are freed from routine tasks and able to focus more profitably on business and commercial strategy.

Today, airlines are confronted by a rising tide of data and competition, which have comprehensively outgunned the industry’s traditional revenue management systems. By harnessing the power of deep learning, FLYR gives airlines the means to sharpen their pricing strategies and achieve sustainable commercial advantage. This extends beyond the basic commercial decisions regarding fare pricing across the entire airline operation, including ancillary offers and pricing, marketing decisions, cargo pricing and capacity, personalized product recommendations, and more.