Cruise Line Revenue Management: Demand Forecasting And Revenue Optimization

The ruise line industry has become one of the fastest growing and most dynamic segments ofthe travel and tourism industry. In recent years, it had experienced an annual growth rate ofover 8% in terms of the total number of cruise passengers. Cruising has become a major part ofthe tourism industry with millions of passengers each year. Like airlines and hotels, it reportsall characteristics of revenue management(RM).Over the past decade, both RM and the cruise line industry have experienced rapid devel-opment. RM techniques have been extensively developed in the airline and hotel industries.However, the cruise line industry has received very limited research attention from revenuemanagement perspective. Particularly, past research has not focused on demand forecastingand evaluation, and dynamic pricing issues for cruise lines.Much of the RM literature has touched the capacity allocation and pricing issues. For effec-tive decisions, the most important condition is the accuracy of demand forecasts. The qualityof RM decisions, such as pricing and capacity control, depends on an accurate forecast to alarge extent.Although demand forecasts are crucial to effective RM, there is limited literatureon forecasting issues in an RM setting. For cruise lines, we could not find any research focus-ing on forecasting issues of RM. Therefore, the main purpose of this thesis is to research onforecasting for cruise line revenue management.This thesis first provides a comprehensive review and introduction of the recent researchesand operations for the cruise line industry, including cruise line operation management, gen-eral marketing research, revenue management, forecasting, pricing and revenue optimization,itinerary design, etc.Then, using data from a major North American cruise company, for cruise line revenuemanagement, this thesis applies a variety of forecasting methods to generate forecasts for non-departed cruises, and apply a variety of probability distributions to test the real distributionsbehind the data.In forecasting section, forecasts of demand of non-departed cruises are generated for finalbookings based on the partial data matrix. The thesis mainly applies a variety of (24) forecast-ing methods, which are divided into three categories (non-pickup methods, classical pickupmethods and advanced pickup methods), to generate forecasts of final bookings for the cruisesthat have not departed at a particular reading point, and also focuses on finding the optimal pa-rameters for each method. A two-stage framework are presented to test alternative forecasting methods and compare their performance.In terms of estimation of demand's probability distributions, a verity of probability distri-butions,such as Normal distribution,Log-normal distribution, Exponential distribution,Poissondistribution, Gamma distribution,Weibull distribution,Rayleigh distribution and Negative Bi-nomial distribution, are tested to determine the real distributions behind the data of total book-ings at cabin type level.The results of forecasting show that, in forecasting final bookings of cruise lines, CP-LR,CP-LLR, CP-MA, AP-ARIMA, AP-MA, CP-ARIMA and AP-ES are most insensitive to whichparameters to use; CP-LR, CP-MA, AP-ARIMA, CP-LLR, AP-MA, CP-ARIMA, AP-LR andCP-ES are most insensitive to which data to use. The results would suggest any one of the 8most accurate and most robust methods: CP-LR, CP-LLR, CP-MA (add), AP-ARIMA (add),CP-ARIMA (add), AP-MA (Add), AP-LR, AP-ES (add), and recommend against using multi-plicative versions of pickup methods, especially when the data is unstable or there is a largenumber of weeks left before departure. In terms of probability distribution, the results offerevidence of a reasonable fit for the Normal distribution and Gamma distribution to the data.In addition to forecasting and estimation, this thesis also discuses the capacity allocationof cabins for cruise lines based on the previous forecasting and estimation results. First of all,normal distribution is used to estimate the demand for each cabin types. At cabin type level,based on the demand estimation, capacity is allocated to each cabin type according to EMSR-aand EMSR-b. At booking horizon level, capacity is allocated to each week before departurebased on the forecasting of booking curve.The methods of dynamic pricing have been widely used in revenue management indus-tries, in which a fixed capacity of perishable products sold in a finite selling horizon in orderto maximize total revenue. As one of the fastest growing segments in leisure travel market, thecruise line industry, which can be regarded as traditional revenue management application,has hardly received any research attention from dynamic pricing aspect perhaps due to thelack of access to proprietary data.Therefore, the author present a method, which has a two-stage price adjustment mecha-nism in each reading period, for dynamically setting and adjusting prices over the selling hori-zon for non-departed cruises through demand learning, and evaluate the method using datafrom a major North American cruise company. In the first stage, the parameters of demandfunction in each reading period is adjusted over time by re-running the regression problemwhen the new data is observed, rather than assumed known in advance. In the second stage,a constrained nonlinear programming is solved to determine the optimal prices for remaining periods. Actually, in each reading period, although all the demand functions and prices are up-dated after new demand data is added to the database, only the price for next adjacent periodis applied to accept future demands. The prices for remaining periods will be updated again.Finally,this thesis considers the duopoly service quality level-price competition in thecruise market with two kinds of consumers: price-sensitive consumers and service quality-sensitive consumers. The characteristics of consumers are re?ected by both the price sensitiv-ity and the service quality sensitivity. First, the utility function for each kind of consumers ndderive the demand function for every firm is established based on the Hotelling model. Then,the duopoly quality-price competition is analyzed under two pricing strategies: uniform pric-ing and discriminatory pricing. The results show that under price discrimination the servicequality is higher than under the uniform pricing strategy;cruise companies set lower prices forprice-sensitive consumers and higher price for quality-sensitive consumers. By comparisons,we conclude that the firms should carry out discriminatory pricing when consumer's trans-portation cost is high, should adopt uniform pricing when consumer's transportation cost islow. Additionally, the proportion of different kinds of consumers has significant impacts onequilibrium prices and service qualities.