Market segmentation methods. Abstract: Segmentation methods

The variety of segmentation tasks and market formation conditions have given rise to many segmentation methods. The purpose of the work is to compile a scientific guide to space modern methods and tools for segmenting consumer markets. Among segmentation methods, the most powerful tool is multidimensional data classification methods. The paper discusses various approaches to constructing multidimensional classification algorithms. When studying socio-economic systems, the vast majority of phenomena cannot be directly measured (mental abilities, personal qualities, tolerance, competence, mobility, political beliefs, etc.). In such cases, it is recommended to use latent models or latent structure analysis. In segmentation practice, you often have to deal with qualitative data. Processing of qualitative data is based on the principles of typologies, which are embodied in computer technologies. Using a combination of quantitative and qualitative methods is often the best solution problems of market segmentation.

qualitative data analysis

latent structure analysis

multivariate data analysis

market segmentation

1. Gottlieb, A.S. Introduction to sociological research. Qualitative and quantitative approaches. Methodology. Research practices: textbook - M: Flinta: MPSI, 2005. - 384 p.

2. Zagoruiko N.G. Applied methods of data and knowledge analysis - Novosibirsk: Publishing House of the Institute of Mathematics, 1999. - 270 p.

3. Kanygin G.V. Tools and methodological principles for analyzing qualitative data // Sociology: methodology, methods, mathematical models. – 2007. – No. 25. – P. 70-98.

4. Kochenkov A.I., Tolstova Yu.N. Ideas are latent structural analysis Lazersfeld in modern sociology // Sociology: methodology, methods, mathematical modeling. – 2003. – No. 16. – P. 125-149.

5. Martyshenko N.S., Martyshenko S.N. Technologies for improving the quality of data in a questionnaire survey // Practical Marketing. – 2008. – No. 1. pp. 8–13.

6. Martyshenko N.S., Martyshenko S.N., Kustov D.A. Tools for developing typologies based on questionnaire data in EXCEL // Academic journal western Siberia. – 2007. – No. 1. – pp. 75–77.

7. Maslak A.A. Measuring Latent Variables in social systems. – Slavyansk-on-Kuban: Publishing Center SGPI, 2012. – 432 p.

8. Omelchenko G.G. Graph-theoretic modeling of market segmentation of service consumers // Economics. Right. Seal. KSEI Bulletin. – 2014. – No. 1 (61). – pp. 145-154.

9. Paul R. Gamble, Anthony Marcella, Alan Tapp Chapter from the book “Marketing Revolution” [Electronic resource]. – http://www.marketing.spb.ru/lib-research/segment/revolution.htm?printversion (date of access: December 2, 2014).

10. Ponizovkin D.M. Construction of an optimal connection graph in collaborative filtering systems // Program systems: theory and applications. – 2011. – T. 2. – No. 4. – P. 107-114.

11. Tatarova G.G. Basics of typological analysis in sociological research: Tutorial/ G. G. Tatarova. – M: Publishing House “New Textbook” 2004. – 206 p.

12. Yadov V.A. Strategies and methods qualitative analysis data // Sociology: methodology, methods, mathematical modeling. – 1991. –No. 1. –S. 014-031.

The main driver for the development of any market is buyers. The set of buyers is heterogeneous. They differ in tastes, desires and shopping motives. This phenomenon has given rise to such an important market research tool as market segmentation, the purpose of which is to develop a flexible policy for interaction with different consumer groups.

Currently, there are many approaches to market segmentation, among which it is difficult for practitioners to navigate. Therefore, the work attempted to compile a scientific guide to the space of modern methods and tools for segmenting consumer markets.

Market segment is a set of consumers characterized by the same type of reaction to the products offered and other elements of the marketing mix.

Market segmentation is a formal procedure based on the application of multivariate statistical analysis methods to the results of marketing research.

When segmenting the consumer market, objective segmentation variables (socio-economic, demographic, geographic) are first used, and then subjective segmentation variables (psychographic, behavioral, personal). From full list Segmentation features are usually chosen to include one or more of the most important ones. A large number of segmentation variables can cause the segment to become too fragmented.

There are two approaches to market segmentation: “a priory” and “post hoc”.

In the first approach, it is assumed that the characteristics of segmentation, the capacity of segments, their number, characteristics, and a map of interests are previously known. That is, it is assumed that segment groups of product consumers in this method have already been formed (“a priori”).

In the second approach (“post hoc”), uncertainty about the characteristics of segmentation and the essence of the segments themselves is assumed. It is necessary to search for signs of segmentation, followed by selection and description of segments.

The variety of segmentation tasks and market formation conditions have given rise to many segmentation methods. Currently, market research has become widespread in practice. following methods segmentation:

Multidimensional classification method;

Correlation segmentation method - “K-segmentation”;

Archetype segmentation methods;

Benefit segmentation method;

Method for constructing a segmentation grid;

Grouping method;

Functional map method;

Segmentation method based on Abel matrices;

Method of segmenting consumers according to their degree of loyalty;

Benefit segmentation method.

Among the segmentation methods that have appeared quite recently are the following methods:

Collaborative filtering method;

Latent model method;

Flexible segmentation method;

Componential segmentation method.

Among segmentation methods, the most powerful tool is multidimensional data classification methods. The use of a multidimensional classification method is called cluster analysis.

Cluster analysis is a set of methods that allow you to classify multivariate observations, each of which is described by a set of initial variables. The goal of cluster analysis is the formation of groups of similar objects, which are usually called clusters.

The classification task must be distinguished from the grouping task. The task of grouping is that the data is divided first by the levels of one characteristic, then by the levels of another characteristic, etc. In contrast to the grouping problem, in cluster analysis the formation of groups of objects (classes) is carried out taking into account all grouping characteristics simultaneously.

To solve the classification problem, it is necessary to introduce the concept of similarity of objects based on observable characteristics. Each class should include objects that have a certain degree of similarity.

To quantify similarity, the concept of metrics is introduced. The similarity between objects will be determined depending on the distance in the selected metric space. If an object described by m features is represented as a point in m-dimensional space, then the similarity of the objects to each other will be defined as the distance in this metric space.

Cluster analysis uses a wide variety of distance measurement methods (metrics). An example of one of the most common similarity metrics is the Euclidean distance:

(1)

where is the distance between the i-th and j-th objects;

The value of the r-th attribute for the i-th and j-th object, respectively;

Where n is the sample size;

Where m is the number of features.

To solve a classification problem, it is necessary to consider the distances between each pair of objects. The distances between pairs of objects are reduced to a similarity matrix. This is a symmetric matrix. Zero values are located along the diagonal of the matrix.

Assessing the similarity of objects using distance measures is convenient when using numerical features. But often there are signs measured on other scales (for example, on a rank scale, or, in general, on a nominal scale). In this case, all features used for classification are reduced to representation in binary code. Let's assume that such a conversion has been performed. That is, each object is described by a vector , each of whose components takes the value 0 or 1.

To measure the similarity of the I-th and j-th objects, we introduce the following frequency notations:

The number of matching single features for both pairs of objects (pairs (1,1));

The number of matching zero features for both pairs of objects (pairs (0,0));

The number of matching single features for the i-th and zero features for the j-th objects (pairs (1,0));

The number of matching zero features for the i-th and single features for the j-th objects (pairs (0,1));

The number of single features in the i-th and single features in the j-th objects, respectively;

The number of zero features for the i-th and zero features for the j-th objects, respectively;

- total number of matching features;

- total number of non-matching characteristics;

The total number of characteristics by which the comparison is made.

Examples of similarity measures in a binary measurement scale include the Rao coefficient (2) and the Hamman coefficient (3):

(2)

Another important quantity in cluster analysis is the distance between entire groups of objects. Let us give examples of the most common distances and proximity measures characterizing the relative position separate groups objects. Let - t-th group(class, cluster) of objects, - the number of objects forming a group, vector - the arithmetic mean of the objects included in the group (in other words - “center of gravity” t-th group), a is the distance between groups and .

The most common methods for determining the distance between clusters are: the “nearest neighbor” method (4), the “far neighbor” method (5), and the method for estimating the distance between centers of gravity (6).

(4)

(5)

(6)

Among all classification methods, the most common are hierarchical agglomerative methods. The main idea of these methods is that in the first step each sample object is considered as a separate cluster. The hierarchical procedure consists of step-by-step merging of the closest classes. The proximity of classes is estimated using a distance matrix or similarity matrix. At the first step, the similarity matrix has dimension . In the next step, when two classes are combined, the similarity matrix is recalculated. The matrix dimension is reduced by one and becomes [ ]. The process is completed in steps when all objects are combined into one class.

The process of combining objects can be depicted as a tree graph (dendrogram). The dendrogram indicates the numbers of the objects being merged and the distances at which the mergers occurred. When identifying classes, they are guided by jumps in the similarity metric on the dendrogram.

Widespread in practice economic analysis received a multidimensional classification method, which is known as the k-means method. It was suggested by McQueen. This classification method belongs to the group of iterative classification methods. There are many modifications of this method. Let's consider the classification algorithm in its original form.

Let there be n objects (observations), each of which is characterized by m features. Observations need to be broken down into given number classes - k.

Step zero. From n points of the studied population, k points are randomly selected. These points are taken as class centers.

Iteration. The set of points is divided into k classes according to the minimum distance to the class centers. You can use any metric to calculate distance. The most commonly used is the Euclidean distance. The class centers are recalculated as the centers of gravity of points attached to the classes.

Examination. If the class centers have not changed during the next iteration, then the classification process ends, otherwise we go to the “iteration” point.

There is a large group of multidimensional classification algorithms based on the application of graph theory. A representative of this group is the Crab algorithm. The FOREL group of algorithms is organized on similar principles.

Let's look at the operating principles of the Crab algorithm. The algorithm begins by finding a pair of points with a minimum distance between them. These points are connected by a graph edge. Then the next closest points, from those not connected, are connected to the already constructed part of the graph. This procedure is repeated until all points are connected by graph edges.

Such a graph will have no loops, and the total length of all its edges will be minimal. A graph with such properties is called a shortest non-closed path (SNP). To split a set of points in a graph into two taxons, the longest edge is broken. The process is then repeated until an adequate class structure is obtained. The Crab algorithm has many modifications. With the further development of the algorithm, the concept of distances and criteria for the quality of partitioning is introduced.

When studying socio-economic systems, the vast majority of phenomena cannot be directly measured (mental abilities, personal qualities, tolerance, competence, mobility, political beliefs, etc.), which led to the emergence of a new concept of “latent” variables. To study latent variables, latent models or latent structure analysis (LSA) are used. This is a fairly wide class of models that have found their application in various areas, including when solving segmentation problems.

Latent structure analysis (from the Latin Latentis - hidden, invisible) is a statistical analysis of empirical data, which allows, based on respondents’ answers to a number of questions, to reveal their distribution according to some hidden (latent) attribute. This trait cannot be measured directly, but the variety of questions used allows us to capture its various manifestations. The method was proposed by the famous American sociologist Paul Lazarsfeld.

The essence of the model proposed by Lazarsfeld was as follows. It is assumed that there is some latent variable that explains the external behavior of respondents. This behavior can be explained by analyzing each person's responses to certain dichotomous survey questions. The latent variable is nominal; the number of its values is known in advance to the researcher. The explanatory power of a latent variable is determined by the fact that it is the cause of the relationship between the observed variables.

The basis of classical latent structural analysis is the fundamental Lazarsfeld axiom of local independence: when the value of a latent variable is fixed, the connections between the observed variables disappear.

Subpopulations of respondents with the same values of the latent variable form latent classes.

There are several types of latent models: models for a continuous latent variable; discrete latent variable models; models for dichotomous traits.

Let us consider a general latent structural analysis model for a discrete latent variable, since in questionnaire surveys more often they deal with precisely such signs.

Let there be m classes of latent variable a and n dichotomous questions. Let's denote:

Probability of answering positively i-th question(proportion of those who answered positively to the i-th question)

The probability of answering positively to the i-th and j-th questions simultaneously.

Probability of answering positively to i, j, k-th questions.

Probability of answering negatively to the i-th question and positively to the j-th and k-th questions

The share of respondents who fell into class m.

The probability of answering positively to the i-th question, provided that you are in class m.

In accordance with the formula of total probability and the axiom of local independence, we will compose the main calculation equations:

where a=1,…, m, and s is a set of indices.

Let us give an example of a system of equations for m=2, n=2

, (9)

where - known, - are not known.

In the general form of the latent class model, the number of equations is , and the number of unknown parameters is m(n+1). Obviously, in order for a system of equations to have solutions, it is necessary that the number of unknowns does not exceed the number of equations, i.e. .

The advantage of latent structural analysis over cluster analysis is that this method allows the inclusion of variables measured on different scales in the analysis.

One approach to estimating latent variables is to use the Rasch model.

Measurements using the Rasch model are the process of transforming initial test data into an interval scale natural logarithms. To calculate latent variables in the Rasch model, the concept of “logit” is introduced.

Logit is a conventional unit that can be easily converted to any other scale. Due to the fact that the Rasch scale is interval, this allows the use of a large number of different statistical analysis procedures. In addition, in the interval scale the zero reference point (0) is not fixed, so indicators in logits are converted to another system of indicators, for example, points using linear transformations. And it is most appropriate to take the average value of the indicators of the observed variables as the starting point in the logit system.

Initially, Rasch models were used to analyze knowledge during testing. Currently, the scope of application of models has expanded to other objects. For example, the apparatus of Rasch models can be used for structural analysis of questionnaire data, including questions characterizing tolerance to various socio-economic processes. Tolerance can be expressed in choosing answers from the list:

Strong agreement (of course yes);

Weak agreement (more likely yes than no);

Weak disagreement (more likely no than yes); strong disagreement (of course not).

The basic equation for a polytomous Rasch model (variables varying at more than two levels) is the equation:

(10)

where x is the gradation of the indicator variable (varies from 0);

Assessment of the j-th respondent for the i-th indicator (questionnaire item);

Probability of the j-th respondent choosing gradation x for the i-th indicator;

Location of the jth respondent on the “effectiveness of socio-economic tolerance” scale (measured in logits);

Location of the i-th indicator on the same scale;

Relative assessment of the l-th gradation;

An index variable that sequentially accepts all answer options for indicative questions.

This model allows you to measure the level of tolerance and the information content of questionnaire items on the same scale (in logits).

One of these methods is presented in the work. This method is based on the application of graph theory.

The considered mathematical model of service market segmentation is based on a 3-partite 3-homogeneous hypergraph. The tops of the first beat, i.e. , one-to-one correspond to the elements of the set of services provided by an enterprise or group of enterprises. Each vertex of the second part uniquely corresponds to some element from the set of consumers classified by demographic characteristics. The vertices of the third share correspond one-to-one to the elements of the set of consumers classified by “income level”: low, below average, average, above average, high, elite class. To construct a set of edges, all possible triplets of vertices are considered - such that . Any such triple is called acceptable if the service can be acceptable to consumers of a given income level and for a given demographic category. The set of all edges is defined as the set of all admissible triples , , .

In a hypergraph, an admissible solution to the problem of service market segmentation under consideration is any subhypergraph , , in which each connected component is a simple star centered at vertex . Through let us denote the set of all feasible solutions to the problem of covering a hypergraph G with stars. For example, one of such solutions is shown in Fig. 1.

Rice. 1. Admissible coverage of the graph by stars

To numerically assess the quality of feasible solutions, three weights are assigned to each edge of the hypergraph, which represent expert assessments. The weights can be: the power of the service in a given positioning, the expected stability of a given positioning, and others. The quality of feasible solutions to this problem is assessed using the vector objective function (11).

(11)

IN lately A segmentation method based on collaborative filtering is becoming increasingly widespread.

In general, the collaborative filtering algorithm can be described as follows: there are many users and many objects. Each user has a list of objects he has rated. Ratings can belong to different scales from 1 to 10, from 1 to 5, etc., as well as different types scales: ordinal or relative. If a user wants to receive a recommendation (or a forecast of his rating for an object that has not been rated by him), then, based on known ratings, the users closest in preferences (or in ratings for the same objects) to the set are established. Next, the algorithm makes recommendations to the user (or calculates a predictive rating for an object), based on the ratings of those closest to the users based on their preferences.

IBM's own concept of value-based segmentation was proposed by the IBM Institute for Business Valuation (Figure 2).

Rice. 2. Segmentation principle used by the IBM Business Valuation Institute

In the tourism business, methods of market segmentation based on psychographic characteristics have become widespread. Market segmentation by psychographic characteristics is the process of dividing all market buyers into homogeneous groups according to criteria such as values, beliefs, motivation to purchase a product and personality type.

Segmentation by psychographic characteristics is based on the theory of typological analysis.

Typological analysis is a meta-methodology of data analysis, a set of methods for studying a social phenomenon that makes it possible to identify socially significant, internally homogeneous, qualitatively different groups of empirical objects, characterized by type-forming features, the nature of which is different, and interpreted as carriers various types existence of the phenomenon.

Typological analysis is mainly based on the analysis of qualitative data. According to V.A. Yadov, qualitative methods allow us to better understand the phenomenon being studied and offer multiple interpretations.

In the world literature, the concept of “qualitative research methods” has been established, which are sometimes called “soft” in contrast to “quantitative” and “hard”. In a certain sense, we can also talk about unformalized or weakly formalized approaches in comparison with strictly formalized ones.

The basis of the typology is a set of judgments (statements) about the proximity (similarity, similarity) of objects, carriers of information about the social phenomena (phenomena, processes) being studied.

The subject of typology is a set of basic characteristics of a social phenomenon, responsible for classifying empirical objects into a group of the same type.

Let's look at some terms that are used in typological analysis.

Type (eng. type) - type, form of existence of social phenomena in science or in the everyday life of people. Type is an entity, knowledge of which is always relative. It has three conventional meanings - typical, typological, typical.

Typological - in typological analysis: special, general, unifying.

Typification (eng. typization) - the construction by people of social reality based on giving labels to others, spontaneous classification.

Typologization is a procedure for systematizing knowledge about the phenomena being studied, either for introducing (assigning) types, or for searching for knowledge about types. Typology serves to construct types.

Typology (eng. typology) - A set of types, the result of their construction. Method of constructing types.

A type-forming feature is a characteristic, a property of social phenomena, on the basis of which types are either constructed or hypotheses about their existence are formulated. A type-forming feature is a conceptual variable.

There are several approaches to solving the problem of unstructured or informal data. The very nature of non-numeric information contains the possibility of using typological analysis to generalize and structure it. “For all the uniqueness of the acting individual, most of his individual meanings are typical, that is, they have commonality with other people.” One of the main tasks is to develop an algorithm for constructing typologies in such a way as to overcome the subjectivity of the researcher without losing sight of important information. Algorithms involve the use of compression and structuring of information so that it preserves the properties of the object under study.

Three approaches to constructing typologies can be distinguished:

The concept of typological operations by A. Barton and P.F. Lazarsfeld;

Analysis of the structure of W. Gerhardt;

Typological analysis of U. Kukartz.

The first approach is based on the use of “typological operations”: reduction, substruction and transformation. By determining the characteristics and degrees of their expression, a space of properties is constructed that underlie the typology. Using graphical or tabular forms of data presentation, all possible combinations and all potential types are determined. All types are associated in one space. Combinations of features can be reduced. Substruction reveals the very space of features that underlies the typology, which can be transformed when interpreting constructed types. Although this approach has been proposed for type construction in quantitative strategy research, it is central to the generalization of non-numerical information.

The second and third approaches use the same basic typological operations, while trying to overcome its main drawback: defining criteria for selecting features for data analysis. The second approach is based on “ideal types”, which serve as the basis for analyzing the information obtained during the study. At the first stage, cases are compared through their reconstruction in order to identify their features. This brings transparency to the study of the generalization process and its results. At the second stage, the cases under study are grouped by comparing them. These techniques generally correspond to the “concept of typological operations” and make it possible to find out all potentially possible combinations of characteristics. At the last stage, semantic connections within and between the resulting groups are identified and explained. For this purpose, a structure and process analysis was developed, consisting of two synthesis steps. The main disadvantages of this approach are the difficulty of abstracting from the subjective views of the researcher when generalizing data and the lack of an algorithm for checking constructed types.

Typological analysis in this approach has a number of features compared to other methods of summarizing qualitative data. In the process of work, they abstract from each individual case under study and obtain a typical event as a result of the ordered phases of the sequence. “Structural hermeneutics,” on the contrary, understands the material in its singularity, inseparability from each specific case under study. Typological analysis is at the intersection between individual history and generally accepted history. In the second approach, the cases under study are preserved, as far as possible, in their integrity; in the third approach, thematic generalization is used when analyzing individual cases and comparing them.

The third direction cannot be fully considered a separate approach. It is “an instrument built for the purpose of expressing the methodological views of M. Weber.” When developing typological analysis tools (software), an attempt is made to combine various ways typology of non-numerical information taking into account their advantages and disadvantages.

Well-known sociologists N. Fielding and R. Lee in the subject area tools For the analysis of qualitative data, it was proposed to use the special term “computer-assisted qualitative data analysis” (Computer Assisted Qualitative Data AnalysiS, CAQDAS). Modern computer-assisted analysis of qualitative data is a methodological research area that brings together scientists from many countries. Assisted analysis is represented by a variety of computer packages, including: Atlas.ti, MAXQDA, NVivo, xSight, Qualrus, Ethnograph, etc. These packages are a class of computer programs that include in their architecture special structures called qualitative coding and reconstruction functions, non-numeric data (coding and retrieval functions). Data coding and reconstruction functions (DCR) are a computer tool (tool) used in human-machine mode and assist the user in studying data presented in so-called non-numeric formats. The basis of assistance is the apparatus of analytical redesignations, called codes, the introduction and connection of which is carried out by the user himself. A more detailed analysis of the methodological developments of computer tools for analyzing qualitative data is presented in the work.

Noting the wide possibilities of foreign computer tools for analyzing qualitative data, it should be noted that they have not found their distribution not only among domestic scientists dealing with the problems of market segmentation, but also among domestic sociologists. These tools have a slightly different focus and are more intended to solve humanitarian and linguistic problems.

The papers present a methodology for analyzing qualitative data that is collected for the purpose of researching markets and, in particular, the tourism market. The methodology includes fairly easy-to-use computer software tools that can be used within the EXCEL computer environment, which is the most common among domestic market researchers. The proposed methodology is based on presenting data in the form of “terms”.

A term is a symbolic expression: , where t is the name of a term, called a functor or "function letter", and are terms, structured or simple. For a formal description of terms, a new concept was introduced in the work - a composite feature.

The proposed methodology for analyzing data used in segmentation should be considered not so much as a doctrine of methods, but rather as a doctrine of the interaction of methods with each other on different classes research practices of data analysis. Structurally, this methodology includes techniques and methods for collecting and measuring information, as well as mathematical methods.

Using a combination of quantitative and qualitative methods is often the best solution to a market segmentation problem. Various methods complement and control each other, the limitations of one method are balanced by the limitations of the other. Such properties are called complementarity and triangulation.

Complementary are dissimilar or even opposite theories, concepts, models and points of view that reflect different views on reality.

Triangulation is the ability to use multiple sources of information. In market analysis, there are several types of triangulation: data triangulation; triangulation of researchers; triangulation of methods; triangulation of theories.

Reviewers:

Latkin A.P., Doctor of Economics, Professor, Director of the Institute for Training Highly Qualified Personnel at Vladivostok State University of Economics;

Embulaev V.N., Doctor of Economics, Professor of the Department of Mathematics and Modeling at VSUES, Vladivostok.

Bibliographic link

Martyshenko N.S., Gracheva V.V. MODERN METHODS AND TOOLS FOR SEGMENTING CONSUMER MARKETS // Contemporary issues science and education. – 2014. – No. 6.;
URL: http://science-education.ru/ru/article/view?id=16405 (access date: 03/31/2019). We bring to your attention magazines published by the publishing house "Academy of Natural Sciences"

The next stage of market segmentation is the selection of a segmentation method and its application. This work is carried out using special classification methods according to selected criteria (features). There are many classification methods, generated by the differences in goals and objectives facing researchers. The most common methods of market segmentation are the method of groupings according to one or more characteristics and methods of multivariate statistical analysis.

Grouping method consists in sequentially dividing a set of objects into groups according to the most significant features. A certain characteristic is singled out as a system-forming criterion (owner of the product, consumer intending to purchase the product), then subgroups are formed in which the significance of this criterion is significantly higher than for the entire population of potential consumers of this product. By successive splits into two parts, the sample is divided into a number of subgroups.

For segmentation purposes, multidimensional classification methods are also used, when separation occurs according to a complex of analyzed characteristics simultaneously. The most effective of them are methods of automatic classification, or otherwise cluster analysis.

In this case, classification schemes are based on the following assumptions. Consumers who are similar to each other in a number of ways are grouped into one class. The degree of similarity among consumers belonging to the same class should be higher than the degree of similarity among people belonging to different classes.

By using similar method the problem of typification is solved with the simultaneous use of demographic, socio-economic and psychographic indicators. The construction of a typology is the process of dividing the studied set of objects into fairly homogeneous and stable groups in time and space.

The largest number of consumers are ordinary buyers. Consumer segmentation based on cluster analysis is a “classical” method. At the same time, there are methods of market segmentation based on the so-called “product segmentation” or market segmentation according to product parameters. She has especially important when releasing and marketing new products. IN modern conditions to increase their competitiveness and correct definition market capacity, it is no longer enough for an enterprise to segment the market in only one direction - defining consumer groups according to some criteria. As part of integrated marketing, it is also necessary to segment the product itself according to the most important parameters for its promotion on the market. For this purpose, the method of drawing up functional maps is used - conducting a kind of double segmentation, by product and consumer.

“Functional maps” can be single-factor (segmentation is carried out according to one factor and for a homogeneous group of products) and multi-factor (analysis of which consumer groups a specific product model is intended for and which of its parameters are most important for promoting products on the market). By drawing up functional maps, you can determine which market segment a given product is designed for, which functional parameters correspond to certain consumer needs.

When developing new products this technique assumes that all factors reflecting the system of consumer preferences, and at the same time the technical parameters of the new product, with the help of which it is possible to satisfy consumer needs, must be taken into account; consumer groups are identified, each with its own set of requests and preferences; all selected factors are ranked in order of importance for each consumer group.

The choice of market segmentation methodology represents difficult task. When an analyst faces the task of market segmentation, he needs to decide on the technology and methods for constructing segments.

After determining the principles and methods of segmentation, the main step before carrying out the segmentation itself is the selection of reasonable criteria for this procedure. Obviously, these criteria will be different for the consumer and industrial markets.

1.3 Segmentation of consumer markets

Consumer market is a market for end consumers who purchase goods for personal, household or family use. Kotler, F. Fundamentals of Marketing. - M.: Rosinter, 1996.

There is no single method of market segmentation. The enterprise needs to test segmentation options based on different parameters(one or more at once) and try to find the most effective approach.

Geographical segmentation .

The market can be divided into different geographical units: states, regions, cities, territories and microdistricts. A firm may decide to operate in one or more geographic areas, or in all areas, subject to the differences in needs and preferences determined by local conditions.

Segmentation by demographics .

It is possible to segment the market into groups based on demographic variables such as gender, age, family size, stage life cycle family, income level, occupation, education, religious beliefs and nationality. Demographic variables are the most popular factors that serve as the basis for identifying consumer groups.

Demographic variables used for segmentation:

1. Age and stage of the family life cycle.

The needs and capabilities of buyers change with age. Even a 6-month-old child is already different in its consumer potential from, say, a 3-month-old. Recognizing this, toy companies are developing different toys for sequential use by their children during each of the months of the first year of life. The focus on a certain age and stage of the family life cycle is not always correct.

Segmentation based on gender has long been applied to clothing, hair care products, cosmetics and magazines. From time to time, the possibility of gender segmentation is discovered in other markets. Most brands of cigarettes are used by both men and women without distinction. However, “female” cigarettes with the appropriate aroma and in appropriate packaging began to appear on the market more and more often, the advertising of which focuses on the image of the femininity of the product.

3. Income level.

An old technique for dividing the market for goods and services such as cars, clothing, cosmetics, education and travel is segmentation based on income level. Sometimes the possibilities of such segmentation are realized in other industries, for example, in the production of alcoholic beverages.

4. Segmentation by several demographic parameters.

Most firms segment their markets by combining different demographic variables. For example, multifactor segmentation can be done based on age, gender and income level.

Segmentation based on psychographic principles . In psychographic segmentation, buyers are divided into groups based on social class, lifestyle, or personality characteristics. Members of the same demographic group can have completely different psychographic profiles.

1. Social class.

Belonging to a social class greatly affects a person’s preferences for cars, clothing, household utensils, leisure activities, reading habits, and choice of retail outlets. Many firms design their products and services with members of a particular social class in mind, providing features and characteristics that appeal to that particular class. Unfortunately, studies of the formation of the class structure of Russian society during the transition period are few.

2. Lifestyle.

It influences interest in certain products and the lifestyle of consumers. Sellers are increasingly resorting to segmenting markets on this basis. For example, the plan is to create jeans for the following groups of men: pleasure seekers, "traditional" homebodies, fidgety workers, "business leaders" or successful "traditionalists". Each group needs a specific cut of jeans, at a different price, offered through different advertising texts, through different trading enterprises. If a company doesn't announce the lifestyle the product is intended for, its jeans may not generate interest.

2. Personality type.

Personality characteristics are also used by sellers as a basis for market segmentation. Manufacturers give their products characteristics that match the personal characteristics of consumers. For example, it has been observed that the personality types of American convertible and hardtop car owners are different. The former are more active, impulsive and sociable.

Segmentation based on behavioral principles. In behavioral market segmentation, buyers can be divided into groups based on their knowledge, attitudes, usage patterns, and reactions to the product. Behavioral variables are considered the most appropriate basis for the formation of market segments.

1. Reasons for making a purchase.

Buyers can be differentiated depending on the reason for the idea of purchasing or using a product. For example, the reason for air travel may be entrepreneurial activity, vacation or family problems. An airline may specialize in serving people who have one of these predominant reasons.

2. Benefits sought.

One powerful form of segmentation is classifying buyers based on the benefits they seek. It was found that in the United States, approximately 23% of buyers purchased watches at the lowest prices, 46% were guided by factors of durability and quality of the product when purchasing, and 31% bought watches as a symbolic reminder of some important event.

3. User status.

Many markets can be broken down into the following segments: non-users, former users, potential users, new users and regular users. Large firms that are looking to gain a larger market share are especially interested in attracting potential users, while smaller companies are looking to win regular users. Potential and regular users require different marketing approaches.

4. Intensity of consumption.

Markets can also be divided into groups of weak, moderate and active consumers of the product. Heavy users typically make up a small portion of the market, but they account for a large percentage of total product consumption. Using the example of beer consumption in the USA, we can see that 68% of respondents do not drink it. The remaining 32% consists of two groups of 16% each: weak consumers (12% of total beer consumption) and active (88%). Most brewing companies focus on active consumers.

3. Degree of commitment.

Market segmentation can also be carried out according to the degree of consumer commitment to the product. Consumers may be loyal to brands, stores, and other distinct entities. Based on the degree of commitment, buyers can be divided into four groups: unconditional adherents, tolerant and fickle adherents, and “wanderers.”

Unconditional adherents -- These are consumers who always buy the same brand of product.

Tolerant Followers -- These are consumers who are loyal to two or three brands.

Fickle Followers -- These are consumers who transfer their preferences from one product brand to another. The pattern of their purchasing behavior shows that consumers gradually shift their preferences from one brand to another.

"Wanderers" - These are consumers who are not loyal to any branded product. The non-committed consumer either buys any brand available in at the moment, or wants to purchase something different from the existing range. Any market is represented by a different combination of buyers of these four types. A brand loyalty market is a market in which a large percentage of buyers demonstrate unconditional commitment to one of the product brands available in it.

6. The degree of readiness of the buyer to perceive the product.

At any given point in time, people are in varying degrees of readiness to purchase a product. Some are not aware of the product at all, others are aware, others are informed about it, others are interested in it, others want to purchase it, and others intend to buy it.

7. Attitude to the product.

The market audience may be enthusiastic, positive, indifferent, negative or hostile towards the product. Experienced political party campaigners making door-to-door pre-election campaigns are guided by the voter's attitude when deciding how much time to spend working with him. They thank voters who are enthusiastic about the party and remind them of the need to vote; they waste no time trying to change the attitude of negative or hostile voters, but strive to strengthen the opinion of those who are positive and win over indifferent ones.

The more clearly the connection between attitudes towards a product and demographic variables is revealed, the more work more efficiently organizations to reach the most promising potential clients.

Submitting your good work to the knowledge base is easy. Use the form below

Students, graduate students, young scientists who use the knowledge base in their studies and work will be very grateful to you.

Basic methods

TO key methods marketers include:

Cluster analysis of consumers. Clustering is the result of the formation of groups of consumers that are united by similar answers to the same question. Buyers are grouped into a cluster based on the same age, income level, hobby or social/family status.
Segmentation by product and sales market stimulation. This method is used before starting the process of creating, developing and producing a new product. An assessment of the capacity of segments can be determined by analyzing and segmenting a niche for a specific product and its characteristics, namely by applying the market summation method.
Working on functional maps. This method is based on a more thorough – double – segmentation of the product itself and the potential consumer of the product.

Let's take a closer look at the first type. To combine groups, you need to use similarities between prospects based on different metrics. When working with taxonomy, so-called clustering algorithms are required, which can well be implemented in the form of hierarchical trees with groups of clients.

One of the most common algorithms is PRIZM. It begins the segmentation process by narrowing down a set of 500 possible social and data points. This system allows you to create a specific segment. As a result, cluster 50, for example, families, is identified. This indicator indicates the level of the most successful professional career, high income and education, as well as the presence of large property and middle age people.

Another example of segmentation based on this method is “consumer attitude towards an updated or new product.” The analysis is carried out based on the characteristic. Several categories of people are identified, including super-innovators, innovators, super-conservatives, conservatives and ordinary buyers. Based on them, you should analyze the niche and identify the true attitude towards new products before their creation and release.

Modern trends in the field of analytics and promotion dictate sophisticated methods of market segmentation. This is necessary for effective analysis of the market capacity of a company or enterprise. Integrated marketing pushes for segmenting the product itself according to the highest priority criteria for its rapid promotion in the market. The characteristic of this method is segmentation by one specific factor or for a similar category of goods.

Integrated marketing pushes for segmenting the product itself according to the highest priority criteria for its rapid promotion in the market.

In addition, it is important to analyze the definition of the target buyer of a product model: what specific criteria and parameters are important for promoting it in the chosen niche. The functional map method allows you to determine which market segment a product is designed for and what parameters and characteristics it should have.

Even in the process of developing a strategy for a new product, this method makes it possible to take into account all the purchasing criteria and preferences of the target consumer.

Thanks to this approach, in the process of developing a new product, you can see and correct shortcomings and create a product for your consumer.

The K-segmentation method aims to find segmentations to potentially select relevant segments. According to this method, there is a certain consumer market, the structure of which cannot be studied, and the selection of segments in accordance with such criteria is impossible. This method can only be used if the company has been operating for more than a year. It is not suitable for new companies, since its main goal is to analyze existing customers, form target groups and identify needs based on existing consumers of goods.

The above methods and methods of market segmentation are the most common and effective. However, there are many more of them. The main purpose of this tool in marketing is to obtain an answer to the question of whether it is possible to identify groups of target consumers. In the event that there are not enough such stable groups, it is recommended to use a mass marketing strategy.