Accurate pricing and revenue forecasting in times of constant change is a very hot topic these days.  

Yes, all historical data became useless as soon as COVID affected demand over a year ago, making forecast and optimization models obsolete in a matter of days.

However, the effect on forecasting goes beyond just this fleeting moment and provides a valuable lesson in how airlines should approach the practice moving forward.  

The New Frontier of Forecasting

Since operations resumed in Latin America, there has been little to no contextual stability, at least for periods lasting longer than a couple of weeks.

The constant introduction of new health measures, migration control procedures, chaotic competitive reactions, different pricing strategies and major capacity changes (just to mention a few ingredients) have been an explosive cocktail.  

This is preventing airlines from having a new baseline in the short and mid-term to get back into their comfort zone and primarily rely on historical data for network optimization.

Despite mentioning the previous facts, having the right pricing and availability every day and adjusting it in real time, is not entirely the responsibility of the revenue management (RM) system optimizer nor is calculating an accurate revenue forecast that can be adjusted on the go.

In this new context, RM teams have had to radically change their daily processes and shift their focus in order to increase flight performance.  

"The constant introduction of new health measures, migration control procedures, chaotic competitive reactions, different pricing strategies and major capacity changes (just to mention a few ingredients) have been an explosive cocktail."

For example, time that used to be invested in strategic flight management (e.g., optimization parameterization, forecasting influences inputs, modeling improvements and flight clustering - among other major time-consuming RM processes) has shifted into improving detection methods on tactical opportunities and feeding new short-term trends into behavioral optimization models that require updating much sooner than ever before.  

Speaking analogously, it has been more important to have the right set of thermometers to rapidly detect changes and act than trying to get the perfect medicine. Not because there might not be one, but because the context changes so fast that you find yourself constantly researching for new alternatives to overcome each new challenge.

Strictly talking about optimization models, most ultra-low-cost carriers (ULCC)/ low-cost carriers (LCC) and hybrid airlines have been using behavioral optimization long before 2020 because of its scalability and efficiency benefits. This clearly gives them an advantage over airlines operating hub-and-spoke networks where O&D optimization complexity can be quite problematic.  

"It has been more important to have the right set of thermometers to rapidly detect changes and act than trying to get the perfect medicine. Not because there might not be one, but because the context changes so fast that you find yourself constantly researching for new alternatives to overcome each new challenge."

Indeed, this last category requires optimization methodologies that are very sensitive to model maintenance and data input quality (ODIF Demand forecast) and generally produced under a Bayesian approach.

This is one of the reasons why Legacy/full-service carriers (FSC) are struggling so much as they adjust to new scenarios, even on domestic non-stop markets.

Executing Accurate Pricing & Revenue Forecasting  

At TARMAC, from an availability perspective, we define accurate pricing as the product of a two-layer analysis that contrasts a behavioral network optimization with a tactical detection process that simultaneously seeks out opportunities and risks that usually last for less than a day.  

This is a key concept to understand in the current context. Behavioral optimization allows airlines to rapidly load and change RASM (revenue per available seat mile) flight objectives, triggered by load factor and yield conditions at different levels of aggregation and translate them into a dynamic availability strategy.

It also allows the RM team to adjust the degree of flexibility they are comfortable with by using ranges of action within the same flight strategy, mostly based on their currently indicated stability of the markets.

Almost simultaneously, multiple tactical processes that consume secondary data like Competitive fares or L2B (Look to book), are overwriting strategic decisions at least twice a day. This second layer is the thermometer we metaphorically mentioned before.

After nearly nine months assisting airlines in the region with their adaptation to the new environment, we have revalued the importance of some of these tools.

The ones that allow airlines to thrive most under constant change are outlined below.  

Real-time connection to the PSS and other live commercial data sources to allow Intra-day optimization triggers

  • Only having a single daily data extraction from the PSS is simply not enough when demand is booking just a couple of days in advance. Real-Time booking data in the RM suite allows airlines to take actions on the fly, well before any batch process dependent information is available.
  • Look-to-book data is useful to spot strengthening trends in customers’ willingness to fly well before passengers make their bookings.  By contrast, these ratios can also detect voids in demand and potential schedule cancellation candidates.  

BI tools with data visualization integrated into RM processes that track specific KPIs to detect trends and take action

  • Pseudo live data visualization and daily processes which ensure that decisions and actions are taken in time have become as important as fighting spill, spoil and dilution since the deregulation act.

Competitive fare data

  • Provide live feeds into the RM System or related monitoring tools available that can trigger an alert or, even better, take a controlled action.

AI and machine learning algorithms are a good way to improve the traditional Bayesian or statistical approach

  • Depending on the situation, retraining ML algorithms don’t need massive amounts of observations to deliver good results. Every few weeks, when the context starts to change again, new and recent historical data sets can drive the relearning process.

Reactive pricing process and fare filing

  • Pricing response time has always been extremely important, but today it is more crucial than ever.
  • Having updated fare structures, along with competitive pricing, conveniently displayed in a single view within the RM suite can enable a significant reduction in response time.

Commercial aviation has undoubtedly gone through its most difficult stretch in its history. But in times of adversity we are often provided with catalysts to make meaningful change and come back stronger. If there is one silver lining, it’s that the industry is swiftly embracing change to adopt new and better processes. The same opportunity lies within pricing and revenue forecasting.