Decision-Support For Customer-Centric Offers

Offer Management In The Customer's Best Interest

Customer centricity is the wave of the future and the reality of the present, and airlines are wisely considering customers’ best interests in devising offer-management structures that are truly effective.

The success of nearly every business is defined by its ability to monetize a deep understanding of its customers through product and service offerings that best meet their specific and, quite often, varied needs. Therefore, the  idea of offering differentiated products to customers has gained traction in a wide variety of businesses over the years.

Airlines have traditionally attempted to implement both types of price discrimination – the practice of charging different prices for the same product (through the use of booking classes and fare rules that cater to different customer segments in the market) and product differentiation (offering different products to different customers).

During the early 2000s, airlines embraced merchandising by creating a catalog of products consisting of a variety of ancillaries intended to satisfy different customer needs and wants, and this merchandising effort emerged into a major revenue stream and driver of profitability.

Today, thinking like a retailer is imperative for airlines, which are part of a competitive customer-centric marketplace with evolving consumer demands. However, true personalization of offers to suit the specific needs and wants of individual customers has not yet become the norm in the airline industry.

Until fairly recently, airlines did not have cost-effective technology in place to take a holistic view of their customers, and that reality somewhat palpably prevented them from truly understanding customer travel and lifestyle preferences.

Utilizing customer data in real time to execute a personalized, seamless travel journey experience requires a robust, multi-channel platform that can deliver personalized offers and services, as well as capture information about the customer interaction.

For most airlines, some of the building blocks of a platform are already in place, although they exist in a fragmented technological and functional environment.

During the past several years, Sabre has invested significant time and resources toward developing a holistic solution to the problem.

Decision-Support Framework

The process of making the right offer to the right customer at the right price can be broken down into four components: understand the customer, design the right product, set the right price and sell effectively.

Personalizing The Offer

Making a personalized offer that can and will connect with the customer at every touchpoint involves a number of decision-making steps. These steps are diverse, both in their scope and in the point at which they are to be applied.

Aside from their inherent complexity, they also have to fit into the existing ecosystem of decision-support systems and the execution platform used by an airline.

The process of making the right offer to the right customer at the right price can be broken down into four components:

  1. Understand the customer,
  2. Design the right product,
  3. Set the right price,
  4. Sell effectively.
Segmentation Process Flow

Segmentation analyzer operates in two phases — calibration and deployment. In the calibration phase, clustering algorithms are run on historical data and clusters are identified, where as in the deployment phase, every new customer is assigned to a predefined cluster.

Understanding The Customer

To design the right products, it is essential to understand and accurately characterize customer behavior.

Segmentation is a way of understanding the behavior of customers in such a way that an airline can recognize both their diversities and their commonalities. The objective is to find certain behavioral patterns and preferences exhibited by groups of customers, and to understand overall customer behavior in terms of these groups.

For decades, airlines have used booking classes that are differentiated by rules such as advance-purchase restrictions, refundability, Saturday-night-stay requirements and so forth to artificially “create” these groups.

Now, an airline can leverage numerous data sources and create flexible methodology combining those sources to bring to life the “360 view” of an airline’s customer base.

The full range of data sources include (but are not limited to):

  • PNR databases,
  • Loyalty-program information,
  • Corporate-sales records,
  • Sales and promotions history,
  • Ancillary-services records,
  • Social-network statistics.

Once data points are aggregated and the data consistency is verified, an airline can apply various clustering algorithms (as implemented in a segmentation analyzer) to optimally group similar entities together and segregate dissimilar entities.

The working of the segmentation analyzer consists of two phases: calibration and deployment.

In the calibration phase, clustering algorithms are run on historical data, and clusters are identified. Then in the deployment phase, every new customer is assigned to a predefined cluster.

The segmentation analyzer provides cluster definitions, in addition to tracking and analyzing capabilities.

Characteristics of each cluster, such as the revenue it provides, the typical advance-purchasing behavior of its population, social activity, trip length, ancillary purchases, etc., can be monitored and compared across different markets and times.

Segmentation helps answer these critical questions:

  • Who are my customers?
  • What actions have those customers taken in the past?
  • What is the purpose of the trip?
  • What kind of ancillaries do the customers buy?
Bundle Design Process Flow

Sabre’s approach to designing bundles involves candidate bundle generation (either automatically from historical data, or through manual analysis) and evaluation (in terms of how well they represent historical customer purchase behavior).

Designing The Right Product

Upon understanding the customers, the next step is to design attractive products for them, incorporating their needs and wants.

An individual ancillary may satisfy only one customer-requirement aspect. So the right offer to a customer may need to be a product bundle comprising multiple ancillaries.

Designing product bundles involves being able to do two things: First, identify patterns in historical ancillary-purchase data; then analyze these patterns for significance and meaning.

A highly effective approach to designing bundles involves:

  • Candidate bundle generation (either automatically from historical data or through manual analysis),
  • Evaluation (in terms of how well the bundles represent historical customer purchase behavior).

Both the generation and evaluation steps can be done at varying levels of sophistication.

By separating the problem into these two steps, it becomes possible for an airline to manually design bundles, as well as have a method of evaluating the bundles’ utility to an airline’s customers.

The simplest form of candidate-bundle generation involves finding frequently co-purchased ancillaries, which is an approach commonly used in market-basket analysis in the retail industry. A more advanced approach is to create bundles that are similar to or can summarize co-purchase patterns across transactions.

Cluster Ancillary Purchase Pattern

To identify the right product to offer a customer, one needs to understand the propensity of a customer to buy an ancillary. This image represents the ancillary purchase patterns among customer segments as observed in the same airline data. The seven clusters on the left represents the number of ancillary purchases made by that segment. The eleven bars on the right represent the ancillaries purchased — the length represents how often an ancillary was purchased, and the width of the line connecting a customer segment to an ancillary represents how often the members of a customer segment bought the ancillary.

More sophisticated approaches involve representing historical transactions in terms of the combination of needs and wants they satisfy, then designing bundles that satisfy commonly expressed combinations.

Evolving these approaches even further can lead to combining diverse “need” items and “want” items in the same bundle, with the purpose of stimulating discretionary spend (such as lounge access or preferred seating) alongside purchase of necessities such as baggage.

The evaluation step itself comprises smaller steps.

At its core, one should define the similarity of a single historical transaction to a single bundle. This can then be aggregated appropriately to measure the extent to which historical purchase data (representing concrete customer behavior) is represented in the proposed bundles, as well as the expressive power of the bundles themselves.

Ultimately, the objective is to balance the simplicity of bundle design against the diversity and complexity of customer needs.

Ancillary Co-Purchases

One approach to designing the right product for a customer is to understand the historical purchase patterns. This graph represents the ancillary co-purchase pattern within a customer segment based on real airline data. The nodes represent an ancillary and an edge is placed between two ancillaries if they are purchased together more than a certain number of times. The size of the node represents how often the product has been purchased, and the thickness of the edge represents how often the ancillaries have been bought together.

Setting The Right Price

Products (individual ancillaries and bundles) need to be priced appropriately to maximize revenue.

A fundamental assumption is that customers have a certain willingness to pay for various ancillaries and bundles of ancillaries, and that this willingness is predicated not just on the price at which the product is offered but also on the context of the offering, meaning the nature and characteristics of the trip and the competitive landscape.

Business customers, for example, may be more willing to pay for WiFi access on the flight so they can work while traveling, and the general population of customers may value extra legroom a lot more in long-haul markets than in short-haul markets.

Traditional approaches to willingness-to-pay estimation involve price-testing: offering different prices to customers and measuring their responses (as expressed in purchase rate) while controlling various other factors that the response is predicated upon.

This methodology requires running an enormous number of tests. So a possible solution to the problem may likely involve an efficient adaptive price-testing mechanism called Thompson sampling, which allows an airline to focus its efforts on prices close to the optimum.

Experiments show that this approach can lead to smaller price tests, and thus less “opportunity” loss incurred through suboptimal price offerings while testing.

The dynamic pricing of airline seats, and the current practice of revenue management in general, derives from the fact that airlines are selling a capacity-constrained perishable product to a diverse customer base whose members have different price elasticities.

The same principle applies to ancillaries and bundles.

An effective proposed solution involves determining optimal prices knowing a customer’s willingness to pay and the capacity constraints that apply to the offer, and extending the current revenue-management solution to include ancillary offers.

Price Testing

To sell a product, one needs to price it optimally. The most popular approach in estimating customer willingness to pay is by conducting price testing.

Selling Effectively

Customer needs with respect to an upcoming trip can be complex, and can be defined by a combination of the trip purpose and pre-existing customer preferences, often gleaned through past behavior.

The key is to associate a customer’s travel pattern and preferences to decision-making when the customer hits “Search.”

Loyalty-program profiles provide basic elements that construct biographic information, cabin preference, previous destinations and meal choices, just to mention a few. However, airlines need to focus on building “personas” for their passengers that transcend the basic elements.

Personas circumscribe profiles and enhance the profiles with “choice” elements based not only on a history of seat selection, but also on optional-service selections as identification markers. This information can be enhanced with additional data sources, such as shopping-session clickstream history and patterns.

Selling effectively encompasses identifying the customer and trip purpose when he or she hits “Search,” and making optimal and personalized decisions on what to offer the customer.

There’s an opportunity to personalize the entire experience through every stage in the shopping session including:

  • Personalized display to show the preferred options for the customer at the top of the list,
  • Personalized product with add-ons of optional services that are dynamically tailored to the needs of the customer,
  • Personalized price, marking the price up or down according to willingness to pay and customer needs,
  • Personalized path to make the customer’s navigation through the booking process simple and fast.

Effective delivery of personalized offers necessitates reservations platforms to match passengers with the services they purchased, and departure-control systems to identify the passenger services purchased and their flight (for fulfillment).

This necessity can be accomplished with a single passenger receipt that combines seat and services purchased.

Today, these are separate processes, on separate validating documents such as electronic tickets and associated electronic miscellaneous documents, respectively.

The execution layer on both direct and indirect channels needs to support every aspect of end-to-end merchandising, from creation and optimization to selling, delivering, servicing, accounting and reporting.

Personalization Engine

The last step of offer management is selling effectively, which involves customizing the product display, product offering and pricing, as well as the path.

Moving Forward

From the outset, the goal of this customer/airline research has been to understand how the airline industry is evolving, and what problems carriers need to solve as they try to better understand and service their customers.

Based on this research as well as product and service design, it has become clear that there are several key components in providing proper decision support for customer-centric offers including:

  • Understanding the customer,
  • Designing the right product,
  • Setting the right price,
  • Selling effectively.

The offer-management platform must be capable of deciphering the likelihood of a particular traveler needing or wanting additional products and services based on the traveler’s individual history and the histories of similar travelers. And it must incorporate insights from what is happening in the marketplace to design the appropriate set of products and services to offer.

Furthermore, the system must determine the appropriate price for the products depending on the context, including the current point in the travel lifecycle, remaining capacity and the ability to cross-sell relevant products.

The resulting solution can very well be a win/win for airlines and customers: significant revenue-enhancement opportunities for airlines, and a more satisfying shopping, booking and travel experience for customers.