|
Article Excerpt
The Customer Equity and Lifetime Management (CELM) solution is based on a decision-support system that offers marketing managers a scientific framework for the optimal planning and budgeting of targeted marketing campaigns to maximize return on marketing investments. The CELM technology combines advanced models of Markov decision processes (MDPs), Monte Carlo simulation, and portfolio optimization. MDPs are used to model customer dynamics and to find optimal marketing policies that maximize the value generated by a customer over a given time horizon. Lifetime value optimization is achieved through dynamic programming algorithms that identify which marketing actions, such as cross-selling, up-selling, and loyalty marketing campaigns, transition customers to better value and loyalty states. The CELM technology can also be used to simulate the financial impact of a given marketing policy using Monte Carlo simulation. This allows marketing managers to simulate several targeting scenarios to assess budget requirements and the expected impact of a given marketing policy. The benefits of the solution are illustrated with the Finnair case study, where CELM has been used to optimize marketing planning and budgeting for Finnair's frequent-flyer program (FFP).
Key words: marketing optimization; loyalty programs; Markov decision processes; portfolio optimization; marketing budget allocation; customer equity; customer lifetime value
History: This paper was received August 2% 2005, and was with the authors 7 months for 2 revisions; processed by Gary Lilien.
1. Introduction
The airline industry has made great improvements in customer relationship management. It is awash in customer data, yet most frequent-flyer programs (FFPs) take a "one size fits all" approach to marketing and service differentiation within a given elite level. Despite technological advances and data abundance, most airlines continue to guess customer value, or use inaccurate models for customer valuation. Moreover, most airlines consider the upper tier of their FFP to be their most valuable customer segment. Yet most of today's FFPs are one-dimensional, concentrating primarily on miles flown or points accrued. Unfortunately, however, elite-level travelers are not necessarily the most profitable, nor may they even be the most loyal. Although they might accumulate the most miles, they may not pay the highest fares and may be very costly to serve (for an extensive general dis cussion of principles and pitfalls of Loyalty Programs, see Shugan 2005).
The emergence of low-cost carriers who have started targeting business travelers has applied more significant price pressure than ever before in the airline industry. The fight for a listing in corporate travel intranets and outsourcing partners' airline options leads airlines to sign extremely lean contracts with their corporate accounts.
With loyalty becoming more of a challenge and price changes a fact of everyday business, it is less evident who the loyal customers are and how much value they leave with an airline's FFP, which delivers rewards based primarily on miles flown, regardless of ticket price.
Finnair, a leading European airline, has offered an FFP called Finnair Plus for many years. As part of its FFP, Finnair conducts numerous marketing campaigns targeting more than 700,000 customers. Each customer is exposed to dozens of campaigns per year. These campaigns have different goals, such as cross-and up-selling, minimizing attrition, points accrual and redemption, and tier upgrade, and are delivered through various channels, such as mailings, in-cabin brochures, the Internet, and magazines. The driving business objective of Finnair was to reduce the costs of the FFP adequately while maximizing the lifetime value of its members.
To achieve these objectives, a team of Finnair marketing managers joined forces with IBM researchers and consultants to define a business transformation process. This included redesigning the marketing strategy around Finnair's FFP and implementing change management processes at several marketing functions, such as campaign management, marketing planning, and multichannel communication. The entire project, which was carried out from February 2003 to March 2005, was executed in three phases:
1. Gain deeper customer insight by deriving finer loyalty and value metrics and more homogeneous and customer profiles.
2. Better understand customer behaviors at various phases of the relationship and the underlying levers that Finnair could act upon at every customer contact.
3. Optimize marketing resource allocation to the FFP by focusing on processes where both cost and revenue can be optimized simultaneously.
The analytical steps underlying these phases can be summarized as follows:
* Introduce advanced value and loyalty metrics and enhance existing customer profiling to capture value, loyalty, and response behavior of customers instead of focusing exclusively on transactions and miles-based segmentations.
* Identify customers' different life cycle phases and dynamics using dynamic programming techniques (Markov decision processes (MDPs)).
* Estimate customer lifetime value and risk (volatility) over variable time horizons by combining MDP models and Monte Carlo simulations to estimate the value-risk profile of customers.
* Optimize the planning of campaign sequences per customer profile to avoid saturation, in an effort to maximize the value of customers over a given planning horizon.
* Optimize marketing budget allocation to balance the value-risk tradeoff of the overall portfolio of customers using portfolio diversification techniques. The remainder of the paper is organized as follows. In [section]2, we provide an overview of the literature and discuss the contributions of our approach. In [section]3, we describe how customer dynamics can be modeled using MDPs. Sections 4 and 5 deal with the estimation of such MDPs and their application details. Sections 6 and 7 address the issue of building an optimal customer portfolio, taking into account the risk of the marketing investment. Section 8 provides an overview of the Customer Equity and Lifetime Management (CELM) technology. Finally, [section]9 summarizes the paper and the business impact of CELM for Finnair. Two technical appendixes provide details of the stochastic model used to optimize customer equity.
2. Literature Review and Contributions
Quantitative approaches to the allocation of marketing resources has recently attracted increased research interest, both in the marketing (e.g., Lilien and Rangaswamy 2003, Gupta et al. 2004, Rust et al. 2004, Rust and Verhoef 2005, Tirenni 2005) as well as in the data mining and statistics communities (e.g., Gelbrich and Nakhaeizadeh 2000, Drew et al. 2001, Pednault et al. 2002, Rosset et al. 2003, Tirenni et al. 2005). There is common agreement that marketing initiatives should be evaluated by measuring their impact on customer lifetime value (Rust et al. 2000, Blattberg et al. 2001, Jain and Singh 2002), i.e., the long-term value generated by a relationship with a customer. Customer lifetime value (CLV) is defined as the sum of the discounted cash flows that a customer generates during her relationship with the company (Berger and Nasr 1998).
In this paper, we focus our analysis on dynamic programming and MDP techniques for CLV maximization (the concept of MDP itself and its application in marketing originated from the catalog industry in the 1950s, Howard 2002). Several approaches to CLV estimation using dynamic programming techniques can be found in the marketing science literature (Bitran and Mondschein 1996, Gonul and Shi 1998, Pfeifer and Carraway 2000, Pednault et al. 2002, Ching et al. 2004). However, most of these approaches present several practical limitations that are usually very important in marketing practice. These limitations are mainly related to the following issues:
* Estimation of robust MDPs when modeling the customer relationship and the effects of marketing actions over variable time horizons. To the best of our knowledge, with the exception of Simester et al. (2006), most of the models found...
|
