Is Machine Learning the Future of Airline Revenue Management?



Great airline revenue management analysts are an invaluable asset to airlines. The 5+% increase in revenue attributed to a sophisticated revenue management system can be difficult to achieve without well-trained and highly experienced analysts who can properly use the system and intervene when appropriate. This means that, potentially, groups of 50 individuals could be responsible for $1-$2 billion in incremental earnings per year at each of the larger airlines – accounting for $20-$40 million per analyst.  (These guys deserve a raise).

As many airlines are now collecting more data than they did just a few years ago, machine learning is fast becoming a major technology in airline revenue management. The question is, to what extent will computers “learn” from experienced analysts and replicate their interventions? At a recent travel conference, easyJet explained how machine learning reverses the normal modelling process:

Instead of:

Data => Computer => Program => Output

This is the normal flow of big data analytics used in most airline revenue management systems. The system takes the data, builds a model around the data, and produces output (demand forecasts by fare and recommended inventory allocations by fare).

With machine learning, we have:

[Data + Output (Analyst Intervention)] => Computer => Program

Here, the big data analytics approach is applied to analyst interventions themselves. Therefore, the system can learn from the analysts and automatically adjust the demand forecasts/allocations without human intervention.

Firstly, of course, we need to acknowledge that all analyst interventions – overriding the big data analytics modeling and recommendations – need to be performed with caution. Academic studies have measured a negative value-add from many analyst interventions. Even experienced analysts can intervene too often, or their adjustments may be too large or too small. The new program must have a way to assess the actual value/benefit of the interventions, or else the machine could be learning poor revenue management practices. Ultimately, machine learning needs to incorporate a way to evaluate each intervention, and in in many cases, will need to learn to ignore or offset interventions that don’t add value.

Let’s review some standard analyst interventions and identify positive learning opportunities – interventions that may add value, and which in a machine learning context, can be replicated by computers.

Overall Demand Adjustment

When market demand changes dramatically, airline revenue management systems may be slow to react because they are designed to change the forecast only when an increase/decrease persists for a period of time; they don’t want to adjust the future based on a one-time “blip.” An analyst may be able to determine earlier whether a change is likely to persist, based on external factors like a schedule change, fare changes, or market-specific news. Intervention can thus speed the model to adapt to the new conditions.

Of course, the model could take schedule and fare changes as inputs and through machine learning apply appropriate adjustments itself. In fact, the model may be better able to quantify the effect of a new competing connect service versus a non-stop and the effect of a three-day sale versus a longer term sale.

Sometimes though, analysts are not actually bringing external market-specific information to the model. They simply judge that the recurrent period of strong bookings represents a significant break from history and that the model is too slow to adjust its forecast. Machine learning - finding a consistent pattern in such interventions - can increase the sensitivity of forecasts to demand changes. Such an adjustment may apply on a macro basis or a micro and flight-specific basis.

Inventory Adjustment

Another common intervention involves directly modifying recommended inventory allocations. Analysts effectively apply business rules to their markets – either implicitly or explicitly.  Example business rules are:

  • If flight is booked at 80% 14 days out, close “S” (a low fare revenue “bucket”)

  • If flight is not booked at 50% 21 days out, open “S”

Although not recommended by most academics since the model has an explicit and sophisticated algorithm-based rationale for opening and closing buckets, this intervention can add value when price fences are not effective or when forecasting is particularly difficult. Machines, of course, can definitely “learn” these to make such rules explicit and make the adjustments automatically.


Fundamentally, machine learning can replicate most airline revenue management analyst interventions make such adjustments more disciplined and consistent. It can insure that changes to the model are appropriate and based on a track record of improved results.There is no doubt, that as it continues to advance, machine learning will play a greater role in revenue management. The lesson here is that it can only go so far, and that’s if human revenue management intelligence, experience and knowledge can be effectively brought into the mix.

Interested in learning more about our revenue management solutions for airlines?


Subscribe Here!