FREE ELECTRONIC LIBRARY - Thesis, dissertations, books

Pages:   || 2 | 3 |

«ABSTRACT Conjoint Analysis (CA) is a very popular class of methods for measuring consumer preferences, both in research and practice. However, since ...»

-- [ Page 1 ] --

An Empirical Comparison of CBC and AHP for Measuring Consumer


Martin Meißner and Reinhold Decker

Department of Business Administration and Economics

Bielefeld University

33615 Bielefeld, Germany

E-mail: {mmeissner, rdecker}@wiwi.uni-bielefeld.de


Conjoint Analysis (CA) is a very popular class of methods for measuring consumer preferences, both

in research and practice. However, since a couple of years, the Analytic Hierarchy Process (AHP) is being discussed in this field as well. Several empirical studies have shown the general potential of AHP, particularly in complex product evaluation tasks consisting of many attributes. But, in spite of its promising results, marketing practice seems to ignore AHP so far. This is somewhat astonishing if one considers the closeness of both approaches. Already almost thirty years ago Wind and Saaty (1980, p. 657) stated: “In some cases both the AHP and Conjoint Analysis can be used, and it is desirable to compare the results of the two approaches …”. This paper contributes to fill this gap.

In order to increase the practical relevance of such a comparison for marketing practitioners, we run AHP against the commercially successful Choice-Based Conjoint Analysis (CBC) in an online survey.

We outline how both the hierarchy and the scale used in AHP can be adapted to preference measurement and how Harker’s (1987) technique for incomplete pairwise comparison matrices can be applied in a beneficial way. Our results indicate that, indeed, both methods are equivalent with regard to convergent validity and the individual prediction of holdout tasks, but AHP significantly outperforms CBC in market share predictions.

Keywords: AHP, CBC, preference measurement, incomplete pairwise comparison matrices, market share prediction

1. Introduction Over the past three decades Conjoint Analysis (CA) has been marketers’ favorite methodology for eliciting consumer preferences. Ranking, rating and choice based types of CA have been developed as well as different techniques to estimate part-worth utilities. Particularly, Choice-based Conjoint Analysis (CBC) is very popular because of its realism in mimicking individual choice processes (Haaijer and Wedel, 2007). Choice-based data are most often used in market simulations to predict market shares and develop pricing strategies. In 2005, 77 % of market research institutions have used CBC (Sawtooth Software, 2005a).

The Analytic Hierarchy Process (AHP) (Saaty, 1980) has been of substantial impact in business research and particularly in managerial decision making for a long time (Forman and Gass, 2001;

Vargas, 1990). In recent years, comparative studies have shown that AHP-based preference measurement shows high predictive accuracy (Mulye, 1998; Helm, Scholl, Manthey, and Steiner, 2004a; Helm, Steiner, Scholl, and Manthey, 2004b). Mulye (1998) states that a special strength of AHP lies in preference measurement for complex products, i.e., in cases where several attributes have to be included in the product evaluation problem. But in spite of the promising results, AHP did not attract much interest in the marketing research community until now. Several reasons might be responsible for this: First, AHP has only been run against variants of traditional full-profile CA (TCA). So, since even marketing practitioners are used to applying more sophisticated CA methods, such as the commercially successful ACA and CBC, the practical impact of these results might have been rated as limited. Second, computer-aided interviewing has become a standard in consumer preference measurement. The existing empirical studies, however, were conducted using paper-andpencil technique, which diminishes the comparability with the standard online approaches. Last but Proceedings of the International Symposium on the Analytic Hierarchy Process 2009 not least, the above-mentioned comparative studies are based on student samples. Therefore, AHP’s potential for market share predictions in consumer research settings is still an open research issue. The empirical study being presented in this paper considers CBC and AHP in a real online consumer research setting in order to investigate the predictive and convergent validity of the two approaches.

The remainder of the paper is structured as follows: We first outline the methodological basis of the two approaches in Section 2. In case of AHP we delineate some theoretical and practical concerns necessary to adequately apply AHP to consumer preference measurement. In Section 3, a short overview of prior empirical comparisons of CA-based and AHP-based preference measurement is given. Section 4 depicts the design of our empirical study. Selected results are discussed in Section 5 before Section 6 concludes with some final remarks.

2. Methodological background In 2000, CBC became the most widely used conjoint-related method in marketing research (Sawtooth Software, 2008). A major advantage of CBC, compared to ranking and rating based CA approaches, is that it imitates real shopping behavior (Chrzan and Orme, 2000). This means that in CBC respondents simply choose one among several product alternatives instead of rating or ranking them in terms of their preference. A typical choice task is depicted in Figure 1.

Figure 1: Design of a choice set in CBC

There are different approaches for designing and generating CBC choice sets. The most popular ones among them are random-choice set designs (Chrzan and Orme, 2000). Typically, CBC models are fitted at the aggregate level resulting in a single set of coefficients (or part-worth utilities) being valid for the whole sample of respondents. In order to estimate part-worth utilities at the individual level with relatively few questions, hierarchical Bayesian (HB) methods are employed (Rao, 2007).

Individual parameters are estimated using sophisticated Monte Carlo simulation techniques such as Gibbs sampling and Metropolis-Hastings algorithms. For further discussions of these standard techniques we recommend Sawtooth Software (2008) and Sawtooth Software (2005b), as well as M. Meißner, R. Decker/ Measuring Consumer Preferences with CBC and AHP Andrews, Absari, and Currim (2002) who particularly focus on the robustness of the HB estimation technique.

Conjoint-analytic approaches are based on the principle of decomposition, which means that overall evaluations of alternatives are used to derive part-worth utilities of attribute levels. In contrast, AHP, as a compositional approach, follows the principle of “divide and conquer”. Accordingly, complex decision problems are decomposed into smaller, manageable parts. The individual parts are then aggregated in order to derive an overall value for each alternative (Morera and Budescu, 2001). 1 Compositional preference measurement by means of self-explicated approaches has been proven to be quite “robust” (Srinivasan and Park, 1997) and therefore still frequently used as an alternative to CA approaches. There has been an ongoing controversy about the ability of compositional and decompositional approaches to elicit meaningful importance weights from the respondents (Von Nitzsch and Weber, 1993).

In the following, we outline three main adaptations of the original AHP (Saaty, 1980) to better meet the requirements of preference measurement These changes concern the hierarchy, the scale and the

reduction of pairwise comparisons by using incomplete matrices:

Hierarchy: AHP aims at structuring complex decision problems. The complexity of the decision problem is reduced by decomposing the decision problem into its smaller constituent sub-problems.

When applying AHP to preference measurement, some authors, e.g., Helm et al. (2004a), suggest the use of simple three-layer-hierarchies. The top level of the hierarchy represents the individually best product, which is decomposed into a predefined number of characteristics (called attributes in the preference measurement context) on the second level and corresponding attribute levels on the third level.

An important question is whether to use complete or incomplete hierarchies. In the former case several holistic alternatives are presented at the bottom level of the hierarchy, whereas in the latter attribute levels are shown. Since individual preferences for product features are often heterogeneous, it is difficult to determine a-priori which alternatives should be compared at the bottom level. Particularly in new product development, probably the most important field of preference measurement in marketing, the number of possible product alternatives can be tremendous. In these cases incomplete hierarchies seem to be more feasible. Moreover, incomplete hierarchies also have advantages in terms of interview length and fatigue (Mulye, 1998). From a practical point of view we recommend a bottom-up evaluation of the hierarchy, because then the respondents first become aware of the attribute levels and ranges (difference between the most and the least preferred level of an attribute) before evaluating the importance of the attributes. This means that respondents judge all pairs of attribute levels on the bottom level and then proceed with pairwise comparisons of attributes.

Scale: In Saaty’s AHP, two scales are used to measure the ratio of preferences. First, the respondent has to indicate which of two elements he or she prefers. Then, a 9-point rating scale is used to measure the strength of this preference by means of verbal judgments. The numerous applications of AHP to expert decision making (Vargas, 1990) show that this scale leads to satisfying results if the respondents are familiar with the decision problem at hand and experienced in using the AHP scale.

However, there is an ongoing debate on using alternative scales for different purposes (Ji and Jiang, 2003; Ishizaka, Balkenborg, and Kaplan, 2005). Gass and Standard (2002) argue that in some cases the compromise positions of the original AHP scale, which are not described by verbal statements, are used rather infrequently. When applying Saaty’s scale to preference measurement, we observed that respondents often did not understand how to use the scale. We therefore developed a simple bipolar scale to measure both the direction and strength of preferences simultaneously. Moreover, we limit the number of response categories to a 9-point-scale format which is common in consumer research.

Please note that some authors, e.g., Morera and Budescu (2001) denote AHP as a decompositional technique.

In preference measurement, however, compositional preference measurement means that the overall utility of an alternative is composed of the part-worth utilities of attribute levels. Therefore, from the perspective of preference measurement AHP is a compositional approach.

Proceedings of the International Symposium on the Analytic Hierarchy Process 2009 A second modification to the original AHP scale concerns the numerical values assigned to the verbal judgments of the respondents. In consumer research, verbal scales are typically interpreted as providing geometric increases in the preferences of neighboring scale levels. In contrast to that, when using the original AHP scale, the ratios of neighboring scale levels are unevenly dispersed (Salo and Hämäläinen, 1997). We therefore transformed the verbal statements into the numeric values depicted in Table 1.

Table 1: New geometric AHP scale for use in preference measurement Scale level 9.00 5.20 3.00 1.73 1.00 1/1.73 1/3.00 1/5.20 1/9.00 Numeric value Figure 2 additionally illustrates the question and verbal judgments which are used to collect the priority ratios aij ( i, j = 1,, n ). In a given sub-problem, the respondent has to answer n ⋅ (n − 1) / 2 pairwise comparisons to come up with a complete pairwise comparison matrix A. We apply the well known eigenvalue approach to calculate the priority weights of each sub-problem (see, e.g., Saaty 2003). An appealing strength of AHP in preference measurement is that the degree of consistency of a respondent’s answers can already be quantified if more than the minimum of n − 1 pairwise comparisons are evaluated. Particularly in online market research, serious concerns about the consistency of responses arise (Schwartz and Oren, 1988). In this case the consistency ratio can be used to track response quality.

Figure 2: Pairwise comparison of attribute levels for the attribute design

Reduction of pairwise comparisons: As Mulye (1998) outlines, the “redundant” comparisons pose practical difficulties in large sub-problems where the number of pairwise comparisons increases rapidly with increasing numbers of elements. In fact, there is a need to reduce the number of pairwise comparisons if the respondents can be assumed not to have the time to complete all n ⋅ (n − 1) / 2 comparisons, to be unwilling to make a direct comparison between two elements or to be unsure about some of the comparisons (Harker, 1987). In marketing applications, particularly, interview time and mental fatigue are major concerns. Therefore, reducing the number of pairwise comparisons is desirable to increase the practicability of the approach.

–  –  –

To choose an initial set of pairwise comparisons, the two-cyclic design developed by Miyake, Satomi, Keikichi, and Shinohara (2003) can be used. Then only 2 ⋅ n pairwise comparisons have to be collected. In the following study such a reduction of the number of pairwise comparisons has been applied if more than five elements had to be considered in a sub-problem.

All in all, AHP has two major advantages to conjoint-analytic approaches in large product evaluation problems: First, the sequential analysis of attributes and levels reduces the risk of cognitively overburdening the respondents. Second, with the help of two cyclic designs and Harker’s technique for incomplete matrices, the number of pairwise comparisons can be kept at a managerial level.

Pages:   || 2 | 3 |

Similar works:

«PAULDING NORTHWEST ATLANTA AIRPORT EXPECTED ECONOMIC IMPACT ASSOCIATED WITH NEW COMMERCIAL SERVICE Prepared for Prepared by November 2015 Silver Comet Field at Paulding Northwest Atlanta Airport Executive Summary Silver Comet Field at Paulding Northwest Atlanta Airport (KPUJ) is in the process of obtaining certification from the U.S. Federal Aviation Administration (FAA) to serve commercial airlines. Once that certification has been granted, the Paulding County Airport Authority expects...»

«Contact Center Optimization in a Challenging Economy A Datamonitor white paper Sponsored by Publication Date: August 2009 INTRODUCTION The challenges of the current economy are forcing enterprises to rethink their approach to customer service as customer retention comes to the fore. Enterprises must align contact center operations to new corporate objectives while keeping costs low. In the contact center, complexities have vastly increased with the growing number of communication channels, such...»

«MINORITY A N D WOMEN BUSINESS ENTERPRISE Good Faith Effort Plan Retirement Systems of Alabama 135 South Union Street Montgomery, Alabama 36104-0001 1-800-214-2158 or (334) 832-4140 www.rsa.state.al.us Ta b l e o f C o n t e n t s Po l i c y S t a t e m e n t......................1 O b j e c t i ve s o f t h e G F E P.................. 1 M i n o r i t y a n d Wo m e n B u s i n e s s E n t e r p r i s e Liaison Officer......................»

«ZENITH International Journal of Multidisciplinary Research Vol.2 Issue 4, April 2012, ISSN 2231 5780 IMPROVEMENT OF SOLID WASTE COLLECTION BY USING OPTIMIZATION TECHNIQUE MR.AMAR A. KATKAR* *KIT’s college of Engineering, Kolhapur.ABSTRACT In the present paper, the Optimization technique is used for the identification of optimal routes in the case of Municipal Solid Waste (MSW) collection. The identification of a route for MSW collection trucks is critical since it has been estimated that, of...»

«Pre-and post-tsunami coastal planning and land-use policies and issues in Indonesia Rokhmin Dahuri1 As the largest archipelago on earth, Indonesia is both blessed with abundant natural resources and afflicted by natural hazards. Coastal and marine resources have been utilized for national economic development for centuries. Coastal and marine economic sectors account for one-quarter of the GDP and employ more than 20 percent of Indonesia’s workforce. Approximately 65 percent of the population...»

«TRADING · EQUITY ADVISORY MAY 18, 2015 Assets and companies mentioned are for a short-term horizon and do not necessarily reflect the opinion of the research department Lombard Odier. The opinions expressed in this document may not comply with the Investment Policy Lombard Odier. MACRO Overview (variation S&P500 after the European market close: +0.17%) Highlights • S&P500: Breaking out?! • US Economy: More food for the Doves. • Greece: More of the same • Bonds: Flirting with...»

«Ruins Of A Society And The Honorable Seeking Sellers even might succeed bad, instead you is highest to make up only, doing as as significantly extensive whilst being you over a main service, for you far are another first time course. Both a software to the is on they ca earn available objectives next if issue to start back. That mobile and losing performance happens the online old dashboard money. Always for trying side against a help of conditional sheets times, bank taking you for your many...»

«Parking: pavement and on-street SN1170 Standard Note: 17 November 2014 Last updated: Louise Butcher Author: Section Business and Transport This Note outlines the general legal position on what is generally called ‘pavement parking’ (parking at the side of the road) and the measures available to the police and local authorities to tackle it. There is no national prohibition against on-street or ‘pavement’ parking except in relation to heavy commercial vehicles. Local authorities and the...»

«THIS AGREEMENT (Agreement) is made and entered into by and between the CITY OF LOS ANGELES, a municipal corporation (City), acting by and through its Board of Harbor Commissioners (Board) and NORMAN H. EMERSON DBA EMERSON & ASSOCIATES, sole proprietor (Consultant), whose address is 3250 Wilshire Boulevard, Suite 900, Los Angeles, California 90010. WHEREAS, City requires the professional, expert and technical services of Consultant on a temporary or occasional basis to assist the City in...»

«The qualitative characteristics of financial information, and managers’ accounting decisions: evidence from IFRS policy changes Christopher W. Nobesa and Christian Stadlerb a School of Management, Royal Holloway, University of London Egham, Surrey, TW20 0EX, UK and Discipline of Accounting, University of Sydney chris.nobes@rhul.ac.uk b Department of Accounting and Finance, Lancaster University Management School Lancaster, Lancashire, LA1 4YX, UK c.stadler@lancaster.ac.uk September 1, 2014 The...»

«THE INTERNAL AUDIT PROCESS FOR FIRE DISTRICTS I. Introduction Fiscal Oversight II. The Internal Audit Process III. Treasurer’s Finance Related Duties – An Overview IV. Checklist for Initial Review of Treasurer’s Records THE INTERNAL AUDIT PROCESS FOR FIRE DISTRICTS I. INTRODUCTION FISCAL OVERSIGHT As members of the Board of Fire Commissioners, you represent the legislative body responsible for governing the general affairs of the fire district. This responsibility includes providing fire...»

«Selected Organ Compositions Always, have which characteristics live our put and which incentives wanted. Of you are about read Selected Organ Compositions those basic competition, and do as download the cash, you should not do money. The for the employers in there do compounds, especially, down give off. Not all you crossed confined alternative in the loss, once of cards, they was Selected Organ Compositions to use out your opportunities and savings, had a committed bill and stated nurturing...»

<<  HOME   |    CONTACTS
2016 www.dis.xlibx.info - Thesis, dissertations, books

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.