Untapped Revenue Opportunities

There Are Many Views Of Where The Industry Is Going. So, Where Are We Going?

A unique choice-based revenue-opportunity model measures the effectiveness of airlines’ revenue management systems and strategies. Among many other capabilities, it identifies opportunities to reduce spill and spoilage and increase revenues and overall profitability, regardless of the revenue management system an airline uses.

Although most airline professionals have a working knowledge of the basic concepts of revenue management, many have never heard of the revenue-opportunity model (ROM). ROM is a valuable post-departure analysis tool for revenue management analysts and executives.

Much like military pilots conduct post-flight reviews to compare their planned versus actual mission performance, revenue management practitioners use ROM for post-departure evaluations of their revenue management performance (considering the joint impact of the revenue management system and manual user overrides).

Using the latest technology, the research team at Sabre Holdings® has developed a unique, choicebased version of ROM, which is implemented for airlines on a consulting basis. It considers a wide range of factors such as:

  • Same-flight up-sell and cross-flight recapture,
  • Competition by market,
  • Planned cabin upgrades,
  • O&D network demands.

These factors are considered when determining optimal overbooking levels and availability by flights (nonstop and connections), fare class and point-of-sale across multiple periods prior to departure. Because choice-based ROM is used on a post-departure basis, it is possible to identify the optimal revenue management controls (with perfect hindsight) to produce maximum future revenue.

Airlines can assess differences between the actual and optimal controls to identify opportunities for revenue management improvements. Post-departure performance reviews based on choice-based ROM provide important insights to shape better revenue management control strategies for future flights.

What Is Choice-based ROM?

Airlines spend significant amounts of money upgrading their revenue management control policy-generation algorithms, but even the best revenue management system will sometimes result in poor control decisions. Therefore, the purpose of choice-based ROM is to determine with perfect “20/20” hindsight how well the airline’s revenue management tools and strategies actually captured the available revenue.

Unfortunately, only a few airlines perform detailed, post-departure analyses of their revenue management controls. Most airlines limit their reviews to traditional financial measures only, such as revenue, yield per passenger kilometer or available seat kilometers, and flight load factors. Revenue, yields and load factors are strongly influenced by many factors besides revenue management controls, including scheduling, sales, distribution and pricing actions.

Since network demand (and revenue) can vary substantially due to other, non-revenue management activities, airlines need new metrics that focus more on revenue management practices. One important metric is the percent achieved revenue opportunity, or PARO.

To compute the PARO value, the choice-based ROM application is used to determine the theoretical improvement potential between two different scenarios:

  1. Maximum revenue (based on perfect overbooking and discount allocation controls with hindsight),
  2. Minimum revenue (which represents no revenue management activity, so all low-valued requests are fulfilled before the higher-valued ones and no overbooking is performed).

The difference between the maximum and minimum revenue is known as the “total revenue management opportunity.” The actual revenue typically falls somewhere in between these two extremes, and the percentage distance in which the actual revenue is situated between the minimum and maximum determines the (PARO). The formula used is: PARO = (actual – minimum) / (maximum – minimum).

For example, if the minimum revenue is US$8 million, the maximum is US$9 million and the actual is US$8.4 million, then PARO = (8.4-8)(9-8) = 40 percent. Higher PARO values represent better revenue management performance (e.g. a 70 percent PARO indicates better revenue management performance than a 40 percent value). The business intent of PARO is to measure both:

  1. The degree of revenue improvement gained specifically from revenue management efforts,
  2. The remaining potential for further revenue improvements.

In most cases in practice, PARO is positively valued (ranging between 0 < PARO < 1). However, in certain cases where the revenue management controls are too restrictive, actual revenue may be less than the minimum, resulting in a negativevalued PARO.

The PARO metric is especially valued by revenue management practitioners because it represents the relative revenue performance within the scope of revenue management analysts to manage (at least theoretically). Thus, it is somewhat less impacted by changes resulting from scheduling, sales, distribution or pricing actions than the morecommonly used yield and load-factor metrics.

Revenue Opportunity Model Business Process

Along with PARO, numerous other revenue management performance metrics can be generated from choice-based ROM, including first-choice demand, spill, recapture, spoilage, flight closure rates and demand-weighted percent open.

Sabre Revenue Opportunity Overall Net Spill By Class

Net Spill By Class An airline wants to see a triangular pattern as indicated in the first graphic, which shows a higher proportion of discount customers being spilled versus higher-valued ones. However, the gaps between the optimal and actual spill rates in the middle-valued fare classes, as indicated in the second graphic, highlight the excessively high closure rates due to either insufficient flight overbooking and/or inaccurate ODF-POS allocations.

Customer-Choice Models And Dependent Demands

Early versions of airline revenue management systems focused on the “independent-demand” assumption — the idea being that once a fare class on a flight was closed for sale, that demand was then lost to the competition, or spilled. In reality, however, there is often a good chance that customers whose first-choice product is not available will rebook an alternate flight and fare on the same carrier, which is known as recapture.

Modern revenue management systems have greatly improved revenue performance because they consider “dependent demands,” which include recapture. There are two types of recapture:

  • Same-flight upsell (to a higher fare),
  • Cross-flight recapture (to an alternate flight).

The choice-based ROM solution offered by Sabre Holdings uses an industry-first approach that considers both types of recapture to find the maximum revenue solution. It does so using customer-choice models (see Dr. Sergey Shebalov’s “Customer-Choice Modeling” article in Ascend, 2013, Issue No. 2).

Customer-choice models are widely employed in other industries and have been successful in airline flight scheduling decision making for more than 20 years. In 2008, choice models were introduced into Sabre® AirVision Revenue Manager (O&D solution), which was implemented for Brazil-based GOL.

The newest version of choice-based ROM uses customer-choice models extensively for demand forecasting and revenue optimization. These models help estimate the first-choice demand for a customer’s preferred flight and fare product from historical sales data. Such estimations can be complicated because historical sales data are strongly influenced by availability (for example, no sales during closed periods) and recapture. Customerchoice models accurately estimate the first-choice demand from historical sales data in cases where products were unavailable for sale, and they allow airlines to predict how demand will shift between different flight and fare products accordingly.

Because customer-choice models are flexible, they support the new choice-based ROM application in a wide variety of pricing environments — the same underlying dependent-demand models are useful in pricing environments with fare restrictions as well as restriction-free or lightly restricted pricing. In addition, the choice-based ROM application is not limited to customers using Sabre AirVision Revenue Manager. Provided the required predeparture sales and availability history information is collected, choice-based ROM can function with any airline’s revenue management system.

How Does It Work?

To use choice-based ROM, historical sales data must first be transformed into demand. Historical sales data are influenced heavily by product availability and related product offerings. To estimate the primary underlying demand from historical sales data, two types of transformations must be applied:

  1. Since the historical sales data doesn’t include demands that were turned away whenever products were unavailable for sales, we need to estimate the true underlying demand. This process is referred to as “untruncation.” Untruncating extrapolates the observed sales to estimate demand in cases where a product was unavailable for sale. The magnitude of the extrapolation required depends on the degree of the closure.
  2. Identification of redirected sales in which customers chose an alternate product because their first-choice product was unavailable (known as “recapture”).
Revenue Management Percent Achieved Revenue Opportunity Metric

Opportunity For Substantial Revenue Increase Results from the revenue opportunity model indicate what percentage of revenue is being captured. In the example above, the airline has realized only 40 percent of the total revenue opportunity and can significantly increase revenue with adjustments to revenue management controls and strategies.

Less sophisticated approaches to demand estimation only consider the first type of transformation, which, when used in isolation, leads to double counting of demand. This occurs because both the original demand for flights during closed periods is counted as spill, and the recapture from those flights to alternate ones is counted as demand as well. However, most observable bookings comprise a mix of first-choice demands, recapture (from same carrier) and capture (from other carrier closures).

When dealing with dependent demands, which have substitution effects across flights and classes due to closures, different, more powerful statistical methods of demand estimation are available. In conjunction with Columbia University and New York University, Sabre Holdings published a paper titled “Estimating Primary Demand For Substitutable Products From Sales Transaction Data” on a new approach, referred to as the multi-flight expectation-maximization (MFEM) method, which uses customer-choice models to handle these untruncation effects under dependent demands. The inputs to the MFEM model are historical sales and availability data by pre-departure periods together with average fare-class values and airline share by market. The market share input is used to gauge the relative attractiveness of competitor offerings compared to the airline under consideration.

The MFEM method produces internally consistent estimates of first-choice demand, spill and recapture for each flight, class, point-of-sale and pre-departure period that is present in the historical data. The estimates are generated via a proprietary method developed by Sabre Holdings that ensures the demand mass balance equation is preserved (in other words, sales + spill = demand + recapture) at both detailed and aggregate levels.

Assessing RM Performance

Overly Restrictive Revenue Management Controls The revenue opportunity model results above indicate overly restrictive revenue management controls are in place, which is creating spill and resulting in lower-than-optimal load factors. Detailed results by market are provided to isolate markets drivers and determine when during the booking period the overly aggressive restrictions are in place. Analysts take action in the revenue management system to ease the controls and reduce spill, thus increasing revenue.

The second step in running choice-based ROM is the performance of flight or networkwide, dependent-demand optimization to determine the revenue maximizing inventory controls by flight, class, point-of-sale and pre-departure period. Techniques for handling dependent-demand optimization are relatively new and have only become available within the past few years. However, they have resulted in revenue improvements of several percentage points compared with previous technologies. One such method is the “sales-based linear program” developed by Columbia University and Sabre Holdings (a paper on this approach titled “A General Attraction Model And Sales-based Linear Program For Network Revenue Management Under Customer Choice” is currently in submission for publication).

The final step in the choice-based ROM process is the revenue management analysis of the results. By comparing the perfect hindsight, model-recommended revenue management controls against the actual ones, analysts can gain insights into potential improved strategies for setting revenue management allocations.

This step is crucial because the choice-based ROM outputs are not utilized in their raw form. Instead, analysts review them to determine the optimal control actions from recent postdeparture flights that are appropriate for future flights (see “Assessing Revenue Management Performance” graphic).

Lessons Learned

A number of airlines using Sabre AirVision Revenue Manager as well as those using different revenue management systems have run the new choice-based ROM prototype. Etihad Airways, in particular, has been using choice-based ROM as part of a comprehensive weekly post-departure revenue management performance review and has evaluated departures across a full year of history.

The Abu Dhabi-based airline generates weekly summary reports and narratives of choice-based ROM actions taken by revenue management analysts in each region and major market, which are reported to executive management.

Actual Improvements To Net Spill Rates

Net Spill Rate Performance Improvements Airlines want to turn away lower-fare passengers in favor of higher-paying passengers. In the example, prior to ROM, the airline was experiencing excessively high closure rates in high-value classes. After ROM, the airline spilled a higher proportion of discount customers versus higher-valued ones.

Sabre Holdings and its customers have developed proprietary views of the detailed O&D, service, cabin, class and point-of-sale choicebased ROM control results including:

  • Flight cabin overbooking results by pre-departure time period,
  • Lowest open class offered by service and predeparture time period,
  • Revenue spilled,
  • Revenue spoilage,
  • Show-up rate analysis.

An important part of the post-departure review is to create customized views suited to the needs of a particular airline. The research team at Sabre Holdings has found that developing filterable and sortable spreadsheet-based data views (or pivot tables) at various levels of aggregation can be quite useful.

Active post-departure performance analysis using choice-based ROM can help improve airline revenue and profitability. Customers have observed sustained year-over-year PARO improvements of up to 10 percent as well as significant reductions in spoilage. Network revenue improvements of 1 percent and profit improvements of more than 5 percent have been attributed to improved revenue management control strategies arising from choice-based ROM reviews.

Clearly, airlines continually strive to improve revenue and reduce spoilage. The choice-based ROM model developed by Sabre Holdings offers valuable insights into the effectiveness of most airlines’ revenue management systems and identifies opportunities for improvement.

Revenue Opportunity Modeling Terminology

First-choice Demand:
Demand for a customer’s preferred flight and fare product (their first choice). Historical sales data are used to estimate demand, but such information must be adjusted to account for availability (no sales during closed periods) and recapture from other products that were unavailable. Advanced statistical models are needed to accurately estimate the first-choice demand from historical sales data in cases where products were unavailable for sale.

Flight Closure Rates:
The percentage of flight departures that were completely closed for sale prior to departure. High flight-closure rates may imply opportunities to overbook more or to limit discount availability.

Net Spill Rates (spill — recapture)/ demand:
Ideally, the net spill rates are higher for lower-valued passengers and less for higher-valued ones. Negative values for net spill rates imply a large amount of upsell from lower-valued classes (or recapture from alternate flights).

Percent Achieved Revenue Opportunity (PARO):
Percentage of total network revenue opportunity a carrier realized on a departure date. Executives using ROM pay close attention to trends in PARO because they indicate whether revenue management performance is improving (or worsening) over time.

Recapture:
Customers whose first-choice product was not available but rebooked on an alternate one with the same carrier. Recapture is often subdivided into same-flight upsell (to a higher fare) or cross-flight recapture (to an alternate flight). Accurately estimating recapture from historical sales and available data requires sophisticated statistical models.

Recapture Fare Ratio:
The average fare value of recapture compared to spill. In practice, it is used to determine the magnitude of the fare increase achieved for passengers who are successfully upsold (recaptured).

Spill:
Turned away demand for a specific flight and fare due to unavailability. Spilled passengers are customers whose preferred product is closed for sale and are either recaptured on an alternate flight and fare or are lost to the competition. Consistently high levels of spill on flights with strong load factors may imply opportunities for price and/or capacity increases.

Spoilage:
Empty seats on flights with spilled demand. Spoilage is an important revenue management performance measure because it is governed by the effectiveness of an airline’s overbooking and discount allocation controls. Lower values of spoilage are better (though in practice some spoilage is unavoidable due to demand uncertainty).

Table of Contents