## Competitive Market Structure and Segmentation Analysis with Self-Organizing Feature Maps (1998)

Venue: | in: Proceedings of the 27th EMAC Conference |

Citations: | 7 - 2 self |

### BibTeX

@INPROCEEDINGS{Reutterer98competitivemarket,

author = {Thomas Reutterer},

title = {Competitive Market Structure and Segmentation Analysis with Self-Organizing Feature Maps},

booktitle = {in: Proceedings of the 27th EMAC Conference},

year = {1998},

pages = {85--115}

}

### OpenURL

### Abstract

The simultaneous treatment of two interrelated and well-known tasks from strategic marketing planning, namely the determination of competitive market structure (CMS) and market segmentation, is addressed via application of the "Self-Organizing (Feature) Map" (SOM) methodology, as originally proposed by Kohonen (1982). In the present paper, some major aspects of the methodological basis of the SOM method are outlined and an SOM-based joint CMS-(preference-)segmentation analysis is illustrated using individual brand choice probabilities derived from diary household panel data. 1. Introduction The analysis of competitive market structure (CMS) represents an important concept in the (strategic) marketing planning process (see, e.g., Wind and Robertson, 1983; Day, 1984; Myers, 1996). In an excellent exposition of contemporary approaches to the determination of CMS, DeSarbo, Manrai and Manrai (1993) describe the primary task of CMS analysis as deriving a configuration of products/brands in...

### Citations

3629 |
Neural Networks: A Comprehensive Foundation (2 nd ed
- Haykin
- 1999
(Show Context)
Citation Context ...nal discrete map of ordered class centers through an adaptive procedure — i.e. a sequential presentation of input data vectors frequently referred to as "network training" (see Kohonen, 19=-=82, p. 139; Haykin, 1994-=-, pp. 414). Subsequently, the adaptive SOM procedure is outlined in more detail. A typical SOM network architecture consists of the following components (see Figure 2 for a graphical illustration of a... |

3242 |
The self-organizing map
- Kohonen
- 1990
(Show Context)
Citation Context ...cations of SOM methodology resulted in some successful implementations for solving a variety of categorization and pattern recognition tasks in speech, image or signal processing (for an overview cf. =-=Kohonen, 1995-=-, pp. 191). Introductions to the methodological foundation of SOM models are provided by Kohnen (1982, 1984, 1995), Pao (1989, pp. 182-196), Freeman and Skapura (1991, pp. 263-289), Ritter, Martinetz ... |

1774 | Introduction to the Theory of Neural Computation - Hertz, Palmer - 1991 |

1306 |
Self-Organization and Associate Memory
- Kohonen
- 1984
(Show Context)
Citation Context ...nction which is responsible for the topological ordering of the class centers in the SOM layer as described below. The SOM algorithm can be briefly described by the following iterative procedure (see =-=Kohonen, 1984): (1) Initi-=-alization of Model Parameters: Weights vectors wij (usually with random numbers within the interval [–0.1;+0.1]), definition of the "neighborhood function" hc(i)c(j)(t), the functional for... |

1093 |
Self-organized formation of topologically correct feature maps
- Kohonen
- 1982
(Show Context)
Citation Context ...ace onto a two-dimensional discrete map of ordered class centers through an adaptive procedure — i.e. a sequential presentation of input data vectors frequently referred to as "network training=-=" (see Kohonen, 1982-=-, p. 139; Haykin, 1994, pp. 414). Subsequently, the adaptive SOM procedure is outlined in more detail. A typical SOM network architecture consists of the following components (see Figure 2 for a graph... |

710 | Cluster Analysis for Applications - Anderberg - 1973 |

404 | Multidimensional Scaling - Cox, Cox - 2000 |

307 | Analysis of individual differences in multidimensional scaling via an n-way generalization of Eckart-Young decomposition - Carroll, Chang - 1970 |

288 | Multidimensional scaling - Kruskal, Wish - 1978 |

210 | Adaptive Pattern Recognition and Neural Networks - Pao - 1989 |

179 | Topology representing networks - Martinetz, Schulten - 1994 |

166 | Neural Network Learning and Expert Systems - Gallant - 1993 |

156 |
The Scree Test For The Number Of Factors
- Cattell
- 1966
(Show Context)
Citation Context ...res against descending sizes of SOM layers provides a useful basis for an examination of alternative topologically ordered cluster solutions (see e.g. in factor analysis the "scree plot" acc=-=ording to Cattell, 1966-=-). Figure 4 contains sequential plots for the heterogeneity (MSSE) and simplicity (MSSIPD) measure used in the study as well as for the sum of both measures (MSSE + MSSIPD). As known from cluster anal... |

129 | A Probabilistic Choice Model for Market Segmentation and Elasticity Structure - Kamakura, Russell - 1989 |

124 |
Artificid neural networks for feature extraction and multivariate data projection
- Mao, Jain
- 1995
(Show Context)
Citation Context ...e beginning and decreasing values with time are recommended. At least the following two fundamentally different types of neighborhood functions can be distinguished (see also Kohonen, 1995, pp. 79 or =-=Mao and Jain, 1995, pp. 302): 1. Discr-=-ete neighborhood function with lateral interactions ("neighborhood set updating"), 2. Continuous neighborhood function ("neighborhood kernel updating"). The choice of the first men... |

89 | Marketing Models - Lilien, Kotler, et al. - 1992 |

87 |
Cluster Analysis in Marketing Research: Review and Suggestions for Application
- Punj, Stewart
- 1983
(Show Context)
Citation Context ...ween pairs of products or brands corresponds to the height of their “lowest common ancestor” node. Such tree structures are typically fit by hierarchical clustering techniques (for overviews see e=-=.g. Punj and Stewart, 1983, Arabie-=- and Hubert, 1994 or Gordon, 1996), which also enables the analyst to explore different “levels” of competition or competitive groups of “higher order”. As a selection of models using hierarch... |

86 | D.M.: Neural Networks: Algorithms, Applications, and Programming Techniques - FREEMAN, SKAPURA - 1991 |

47 |
A review of hierarchical classification
- Gordon
- 1987
(Show Context)
Citation Context ... height of their “lowest common ancestor” node. Such tree structures are typically fit by hierarchical clustering techniques (for overviews see e.g. Punj and Stewart, 1983, Arabie and Hubert, 1994=-= or Gordon, 1996), which-=- also enables the analyst to explore different “levels” of competition or competitive groups of “higher order”. As a selection of models using hierarchical tree structures for the determinatio... |

28 |
The PARAFAC model for three-way factor analysis and multi-dimensional scaling
- Harshman, Lundy
- 1984
(Show Context)
Citation Context ...y for dichotomous data structures (see, e.g. Christoffersson, 1977 or Muthén, 1989) and models for three-mode (i.e. simultaneous Q- and R-type factoring) factor analysis (see Kroonenberg, 1983, 1984;=-= Harshman and Lundy, 1984-=-; for marketing specific applications cf. the works of Cooper, 1988; Cooper and Inoue, 1996; Cooper, Klapper and Inoue, 1996). Decompositional approaches are usually based on multidimensional scaling ... |

28 |
Marketing Engineering: Computer-Assisted Marketing Analysis and Planning
- Rangaswamy
- 1998
(Show Context)
Citation Context ...ing research (the latter are represented in marketing textbooks such as Bagozzi, 1986, pp. 244; Moore and Pessemier, 1993; Urban and Hauser, 1993, pp. 253; Lilien, Kotler and Moorthy, 1992, pp. 74 or =-=Lilien and Rangaswamy, 1998-=-, pp. 95). The measurement rules associated to these models regularly correspond to consumers' evaluative processes involving perceptions and/or preference formation and thus ultimately rely on at lea... |

26 | A Clusterwise Regression Method for Simultaneous Fuzzy Market Structuring and Benefit Segmentation - Wedel, Steenkamp - 1991 |

25 |
A simultaneous approach to market segmentation and market structuring
- Grover, Srinivasan
- 1987
(Show Context)
Citation Context ...he average set contains about three brands) lead to a relatively large number of zero-entries. Nevertheless it is widely used practice in CMS analysis (see e.g. the works of Rao and Sabavala, 1981 or =-=Grover and Srinivasan, 1987), it shou-=-ld be noticed that aggregation of choice probabilities over time assumes a homogeneous brand choice process of "zero-order", i.e. stability of the market. 5.2. Properties of alternative SOM ... |

23 |
Three-mode principal component analysis: Theory and applications
- Kroonenberg
- 1983
(Show Context)
Citation Context ...actor analytic methodology for dichotomous data structures (see, e.g. Christoffersson, 1977 or Muthén, 1989) and models for three-mode (i.e. simultaneous Q- and R-type factoring) factor analysis (see=-= Kroonenberg, 1983-=-, 1984; Harshman and Lundy, 1984; for marketing specific applications cf. the works of Cooper, 1988; Cooper and Inoue, 1996; Cooper, Klapper and Inoue, 1996). Decompositional approaches are usually ba... |

22 | Correspondence analysis: Graphical representation of categorical data in marketing research - Hoffman, iianke - 1986 |

21 |
An empirical analysis of the relationship between brand loyalty and consumer price elasticity. Marketing Science 10
- Krishnamurthi, Raj
- 1991
(Show Context)
Citation Context ...(national/diet) 8.1 Du darfst (national/diet) 6.7 Bellasan (private/normal) 5.4 Bellasan Diät (private/diet) 3.0 Others 4.2 Analogous to similar approaches to CMS (e.g., Bucklin and Srinivasan, 1991;=-= Krishnamurthi and Raj, 1991; -=-Balakrishnan, Cooper, Jacob and Lewis, 1996), households’ (unconditional) brand choice probabilities are used as input data for SOM-training. Thus, each individual brand-switching household k may be... |

18 | MULTICLUS: A new method for simultaneously performing multidimensional scaling and cluster analysis - DeSarbo, Howard, et al. - 1991 |

18 |
Segmentation and positioning for strategic marketing decisions
- Myers
- 1996
(Show Context)
Citation Context ...data. 1. Introduction The analysis of competitive market structure (CMS) represents an important concept in the (strategic) marketing planning process (see, e.g., Wind and Robertson, 1983; Day, 1984; =-=Myers, 1996-=-). In an excellent exposition of contemporary approaches to the determination of CMS, DeSarbo, Manrai and Manrai (1993) describe the primary task of CMS analysis as deriving a configuration of product... |

14 |
Cluster analysis in marketing research
- Arabie, Hubert
- 1994
(Show Context)
Citation Context ...r brands corresponds to the height of their “lowest common ancestor” node. Such tree structures are typically fit by hierarchical clustering techniques (for overviews see e.g. Punj and Stewart, 19=-=83, Arabie and Hubert, 1994 or Gord-=-on, 1996), which also enables the analyst to explore different “levels” of competition or competitive groups of “higher order”. As a selection of models using hierarchical tree structures for ... |

13 |
Marketing strategy: New directions for theory and research
- Wind, Robertson
- 1983
(Show Context)
Citation Context ...s derived from diary household panel data. 1. Introduction The analysis of competitive market structure (CMS) represents an important concept in the (strategic) marketing planning process (see, e.g., =-=Wind and Robertson, 1983-=-; Day, 1984; Myers, 1996). In an excellent exposition of contemporary approaches to the determination of CMS, DeSarbo, Manrai and Manrai (1993) describe the primary task of CMS analysis as deriving a ... |

12 | Using a variety-seeking model to identify substitute and complementary relationships among competing products - Lattin, McAlister - 1985 |

11 | Customer-oriented approaches to identifying product-markets - Day, Shocker, et al. - 1979 |

11 | Non-spatial tree models for the assessment of competitive market structure: An integrated review of the marketing and psychometric literature - DeSarbo, Manrai, et al. - 1993 |

11 | A model for the analysis of asymmetric data in marketing research - Harshman, Green, et al. - 1982 |

11 | Market Definition and Segmentation Using Fuzzy Clustering Methods - HRUSCHKA - 1986 |

11 |
Explanatory and Predictive Models of Consumer Behavior
- Roberts, Lilien
- 1993
(Show Context)
Citation Context ...esses involving perceptions and/or preference formation and thus ultimately rely on at least one explicit explanatory variable steming from some reasonable theoretical framework of consumer behavior (=-=Roberts and Lilien, 1993-=-, pp. 43). 1 Vienna University of Economics and Business Administration, Department of Retailing and Marketing; contact address: Augasse 2-6, A-1090 Vienna, Austria, Europe; email: reutterer@wu-wien.a... |

11 | Multidimensional scaling and its applications - Wish, Carroll - 1982 |

10 |
Competitive Maps: The Structure Underlying Asymmetric Cross Elasticities
- Cooper
- 1988
(Show Context)
Citation Context ...sitioning models incorporating actual customers’ purchase information, background characteristics and/or reactions to marketing mix variables (see, e.g., DeSarbo and Rao, 1986; Moore and Winer, 1987=-=; Cooper, 1988-=-; a review is provided by Green and Krieger, 1989). In the present paper, however, the issue of a compact representation of competitive relationship patterns among brands is brought into focus. CMS an... |

10 |
Constructing MDS joint spaces from binary choice data: A multidimensional unfolding threshold model for marketing research
- Hoffman
- 1987
(Show Context)
Citation Context ...ndividual (consumer specific) differences in a “common” product space. Recent contributions comprise internal unfolding models for (binary) brand switching data (e.g. Holbrook, Moore and Winer, 19=-=82; DeSarbo and Hoffman, 1987-=-). In the GENFOLD2 model by DeSarbo and Rao (1986), product coordinates are related to marketing mix variables and product features while consumers' ideal points are related to their (demographic, soc... |

10 | Y.Yu and B.Bavarian, Analysis of the convergence properties of Topology Preserving Neural Networks - Lo - 1993 |

10 | A probabilistic model for the multidimensional scaling of proximity and prelerence data. Marketing Sci - MacKvy, Zinnes - 1986 |

10 |
Design and marketing of new products, 2ndedition
- Urban, Hauser
- 1993
(Show Context)
Citation Context ...ity have been proposed and some of them nowadays are standard tools in marketing research (the latter are represented in marketing textbooks such as Bagozzi, 1986, pp. 244; Moore and Pessemier, 1993; =-=Urban and Hauser, 1993-=-, pp. 253; Lilien, Kotler and Moorthy, 1992, pp. 74 or Lilien and Rangaswamy, 1998, pp. 95). The measurement rules associated to these models regularly correspond to consumers' evaluative processes in... |

9 | Comparative performance of the FSCL neural net and K-means algorithm of market segmentation - Balakrishnan, Cooper, et al. - 1996 |

9 |
Strategic marketing planning: The pursuit of competitive advantage
- Day
- 1984
(Show Context)
Citation Context ...hold panel data. 1. Introduction The analysis of competitive market structure (CMS) represents an important concept in the (strategic) marketing planning process (see, e.g., Wind and Robertson, 1983; =-=Day, 1984-=-; Myers, 1996). In an excellent exposition of contemporary approaches to the determination of CMS, DeSarbo, Manrai and Manrai (1993) describe the primary task of CMS analysis as deriving a configurati... |

9 | Latent Class Multidimensional Scaling: A Review of Recent Developments - SARBO, MANRAI, et al. - 1994 |

9 | Intensity measures of consumer preference
- Hauser, Shugan
- 1980
(Show Context)
Citation Context ...al attributions of product/brand characteristics as perceived by consumers (profile data) into lower-dimensional spaces and employ traditional methods of data reduction, such as principal components (=-=Hauser and Shugan, 1980-=-) or discriminant analysis (Pessemier and Root, 1973). Since standard principal components analysis is not suited for non-metric profile data, Hoffman and Franke (1986) address this problem by using c... |

8 | Recent Contributions to Optimal Product Positioning and Buyer Segmentation,” European - Green, Krieger - 1989 |

8 |
A Consumer-Based Methodology for the Identification of New Product Ideas
- Shocker, Srinivasan
- 1974
(Show Context)
Citation Context ...982). Developments important in marketing research are the (external) multidimensional unfolding extensions, like PREFMAP (for a compact description see, e.g. Green and Rao, 1972, pp. 214) or LINMAP (=-=Shocker and Srinivasan, 1974), w-=-hich allow the representation of consumer preferences as vectors or ideal points together with brands/products in one and the same geometric space (“joint space”), or the INDSCAL model by Carroll ... |

7 |
Simultaneous visualization and clustering methods as an alternative to Kohonen maps
- BOCK
- 1997
(Show Context)
Citation Context ...hm the learning rate α(t) is forced to decrease with time, a classification of the input data space would be achieved via stochastic approximation of the minimum (inner group) variance partition (cf.=-= Bock, 1997-=-).sHowever, as the formulation of the weights updating rule in step (4) of the above algorithm suggests, there is another feature of SOM learning which differs from other clustering techniques: Not on... |

7 |
A Constrained Unfolding Methodology for Product Positioning
- Rao
- 1986
(Show Context)
Citation Context ...rtunities, CMS analysis is extended towards positioning models incorporating actual customers’ purchase information, background characteristics and/or reactions to marketing mix variables (see, e.g.=-=, DeSarbo and Rao, 1986-=-; Moore and Winer, 1987; Cooper, 1988; a review is provided by Green and Krieger, 1989). In the present paper, however, the issue of a compact representation of competitive relationship patterns among... |