Multiattribute Model in Consumer Behaviour Research

ABSTRACT - The techniques for measuring the significance of the product's attributes become more and more important. Taking into account the advantages but also limitations of the full-profile methods the objective of this paper is to analyze the contribution of the multiattribute linear model in measuring the impact of the product's attributes on the market share in new market circumstances. In the paper the model is described and applied in the empirical study in which 43 different cookers being available on the Slovenian market are considered. The approach itself aswell as some results obtained are presented in the paper.


Majda Bastic (1995) ,"Multiattribute Model in Consumer Behaviour Research", in E - European Advances in Consumer Research Volume 2, eds. Flemming Hansen, Provo, UT : Association for Consumer Research, Pages: 398-405.

European Advances in Consumer Research Volume 2, 1995      Pages 398-405


Majda Bastic, University of Maribor


The techniques for measuring the significance of the product's attributes become more and more important. Taking into account the advantages but also limitations of the full-profile methods the objective of this paper is to analyze the contribution of the multiattribute linear model in measuring the impact of the product's attributes on the market share in new market circumstances. In the paper the model is described and applied in the empirical study in which 43 different cookers being available on the Slovenian market are considered. The approach itself aswell as some results obtained are presented in the paper.


The era of automation and higher real per capita incomes have resulted in a number of significant changes in the market place. Some of the most important changes are:

- Market life cycles for products are getting shorter, consequently new designs must follow one another more frequently.

- The market place is demanding a greater variety of prod ucts without increasing the volume desired. This means that a factory must produce smaller quantities of each product.

- The market place is time sensitive. It wants its products on time. This time-based competition means that factories must produce in smaller batches on time-controlled schedules.

- The market place is cost-sensitive. It wishes to lower break- even points. Hence, highly efficient production capabilities are requested with high quality and reliability.

To compete successfully on these markets a company will have to offer an increasing number of new and improved products to the customers, designed in accordance with their requirements and expectations.

The improvement of the quality of existing products and the development of new ones will be among the most important strategic decisions where the "the voice of the customer" will play a significant role. According to the definition of Griffin and Hauser (1993) the voice of the customer is a hierarchical set of customer needs where each need (or set of needs) has assigned to it a priority which indicates its importance to the customer. Developing products based on the voice of the customer becomes a key criterion in total quality management.

Taking into account the experiences of the best a good product is crucial but it is only the beginning. The product must be constantly developed and improved. If the product wants to succeed it must distinguish itself favorably on attributes that are important to a target market segment.

A very important question in the decision-making process for a product improvement is, what are the attributes of the product or service that create value for the customer? And which attributes are most important? The attributes go well beyond physical characteristics and encompass all the support activities and systems for delivery and service. Each market segment has unique attributes that customers employ to judge the competitive offerings.

The strategic segments of a market are formed from distinct groups of products sold to distinct groups of consumers and described by those few segmentation variables that account for large differences in consumers behaviour.

Not all parts of the market have an equal attractiveness. Instead of homogeneity there are diverse submarkets within the market that vary widely because customer groups have different needs and behaviour and consequently the profitability of different submarkets vary too. The immediate payoff from successful segmentation of a market into submarkets is the identification and nurturing of product groups and customer groups where profitability prospects are superior considering the firm's resources and competition.


The techniques for measuring the consumers' trade-offs among multi-attributed products become more and more important. Conjoint method defined by Green and Srinivasan (1990) "as any decompositional method that estimates the structure of a consumer preferences, given his or her overall evaluations of a set of alternatives that are prespecified in terms of levels of different attributes" has received a considerable academic and industry attention in the early 1970, and it continues to be popular for reasons mentioned earlier.

The dominant way to map the product attributes is by asking consumers to indicate "how important" they consider the various product attributes on a set of ordered scales. Green and Srinivasan (1978) described several advantages and also limitations of the full-profile methods in comparison with the trade-off procedure (two-factor-at-a-time method). An important advantage of the full-profile method is its ability to measure overall preference judgements.

Green and Srinivasan (1990) have found out that the full profile method of conjoint analysis works very well when there are only a few (say six or fewer) attributes. Green (1984) indicated the similar findings that industrial users of conjoint analysis have strained the methodology by requiring larger numbers of attributes and levels within attributes, thus placing a severe information overload on the respondents. When faced with such task, respondents resort to simplifying tactics and the resulting part-worth estimates may distort their true preference structures (Wright, 1975). To solve this problem, the average commercial study has used 16 stimuli evaluated on eight attributes at three level each. Such a design leads to no degrees of freedom (Wittink and Cattin, 1989).

Taking into account all that, and adding the increasing number of segments within the product class an interesting question arises: What is the contribution of mathematical models to determining the benefit segments and to measuring the significance of the product's attributes within the segment in new market circumstances?

There are basically two groups of methods which can be used for determining the influence of individual parameters on the value of the chosen criterion; namely mathematical and statistical methods. Following the literature of Horsky and Rao (1984), Shocker and Srinivasan (1979) we can find that estimates obtained from linear programming models show greater stability and greater prediction ability.

For this reason the model proposed by Oral and Kettani (1989) will be applied in this study. Oral and Kettani have introduced some improvements into the Shocker and Srinivasan model which are especially important for solving a product improvement problem. The intensity of customers preferences in terms of market shares and the dependence between the attribute weights and its scores are included in the Oral and Kettani model.

When facing a purchase decision, customers assign utilities to each product and then select the one with the highest utility. Therefore, our approach rests on the assumption that the product's market share can be taken as the measure of the product's market success or failure because it does or does not possess the appropriate values of the desired attributes. Considering the requirements of the Oral and Kettani model the competing products must be ranked according to their market shares. The rank order of competing products is given by the set R={1,...,n} where the rank one has the product with the highest market share and the rank n the product with the lowest one. The products' market shares are given by the vector M=(MI,---,Mn) where Mr denotes the market share of the r-th ranked product.

It has long been acknowledged that product attributes do form hierarchies (Grunert, 1989). To find out the values of weights belonging to individual attributes, attributes can be broken down into attribute values, and this again can occur on various levels of specificity. The attribute price can be broken down into the values high, low, and reasonable, but could also be broken down into numerical values.

Let C be the set of m product attributes under investigation and Vjp be the value of the p-th attribute belonging to the j-th competing product. The weights are assumed to be step functions of attributes values. The range of all values which attribute p can take is divided into non-overlapping and exhaustive smaller intervals that each attribute value falls in one and only one of the intervals. Variable wip expresses the weight assigned to the values of the p-th attribute from the i-th interval.

Taking into account the different nature of individual attributes, wip is

1. non-decreasing step function of Vp, implying wi+l,p>wip for some attributes say p0Cq, Cq{1,2,...,p} or

2. non-increasing step function of Vp, implying wi+ 1,p<wip for the other attributes say p0Cm, Cm={p+1,...,m}.

The nature of the individual attribute helps us to choose the appropriate step function. For example the non-decreasing step function is appropriate function for the weights belonging to values of the attribute quality because it is expected that a higher quality has a greater influence on the market share than the lower one. Therefore, the value of the weight, belonging to the higher value of the attribute quality, must not be smaller than the value of the weight belonging to the lower value of quality attribute. In most cases a lower price contributes more to the market share than a higher one, therefore the non- increasing step function is taken for the weights belonging to individual prices.

Decompose the product's market share into the contributions to the market share obtained by the individual product's attributes. The contribution of the p-th attribute with its value of Vsp is equal to wipVsp, where wip expresses the unknown weight assigned to the values of the p-th attribute from the i-th interval. The theoretical market share of the s-th ranked product is then equal to

Us=SwipVsp forVspE[vip,vi+1,p) sER   (1)

market share Ms, s=1,...,n, will be minimal. The deviations between the theoretical market shares defined by (1) and actual ones given by vector M are defined by constraints

Us-Us+1+Zs+-Zs-=Ms-Ms+l s=1,...,n-1   (2)

where Zs+ is positive if Ms-Ms+l>Us-Us+l and zero otherwise. ZS- has a positive value if US-Us+l>MS-Ms+l and zero otherwise.

The nature of the step function is considered in the model by constraints

Wi+1,p > Wip i=1,....,Ip-1 , PECq   (3)

for non-decreasing step function and by

Wi+1,p < wip i=1,....,Ip-1 , PECm   (4)

for non-increasing step function.

Ip is the number of non-overlapping, exhaustive intervals of all values which attribute p takes.

To obtain logically consistent weights the constraint (5) is added


which assures that the sum of the market shares of all competing products under investigation is equal to unity.

Considering the objective of optimization - find those values of weights that deviations defined by (2) will be minimal - the objective function can be written in the form


The optimal value of the variable wip for the k-th market segment will be denoted by wipk * and is used for determining the significance of the attributes in achieving the market share on that segment. The contribution of the p-th attribute to the market share of the j-th product on the k-th market segment is equal to


and is displayed with other contributions in Table 1.

Data displayed in Table I can be used in the analysis of revealing the sources of the product's strength and weakness on the k-th market segment, which are important data in decision making on the product's improvement or development of new products. Denote by mpk the highest contribution to the market share achieved by the p-th attribute on the k-th market segment. It is equal to


The difference between the values mpk defined by (8) and mjpk defined by (7) reveals the source of product's weakness or

We want to find those values of variables wip so that the strength expressed in lost market share. In Table 2 the calculation difference between the theoretical market share Us and actual of the lost market shares djpk is presented. The value zero in the p-th column belongs to the product with the best value of the P-th attribute and the highest one to the product with the worst value of that attribute on the k-th segment.



Market segments differ in their requirements, response profile, the cost to serve them and in their attractiveness, which can be measured by the size and the growth of the segment, as well as the competition. The size and especially the growth of the segment increase its attractiveness while the increasing number of competitors diminishes it.

The attractiveness of the segment can be put into the model with the coefficient of attractiveness. The coefficient with the value one belongs to the most attractive segment. Taking into account the coefficient of attractiveness of the k-th segment, denoted by ak, the relative loss of the j-th product's market share due to the inappropriate value of the p-th attribute on the k-th segment is equal to

ak(mpk-mjpk) = akdjpk p= 1,...,m k=1,..., r

The relative losses of the market shares on r target segments for the j-th product are shown in Table 3.

The attribute with the highest value of Ljp should be first improved if the desired improvement could be achieved with the existing resources. In many cases the improvement of the product's total quality will not be possible without investments in

- advanced technology

- training people

- new organization

- new materials.


The objective of this study is to analyze the ability of the improved model defined by (l)-(6) in measuring the significance of the product's attributes and marketing instruments on the product's market success as well as to apply the obtained results in yielding insights into consumer behaviour.

The described model has been applied in the research of the consumer behaviour in purchasing cookers produced in Slovenia and competing with other similar cookers produced in Germany and Italy being available on the Slovenian market. 43 different cookers were included in this empirical study. Technological progress linked to the development and production of this product has enhanced the number of product segments, and the attributes as well as the number of levels within the attributes. Taking into account all that, three segments are defined at the beginning:

- cooker

- built-in cooker consisting of

- built-in oven and

- built-in hob.

The crucial task in creating the model is the selection of attributes which will be considered in the model. Grunert (1989) mentioned that the elements of cognitive structure, relevant for the consumer behaviour, are product uses, product alternatives, and product attributes. When attributes are linked to uses, these associative links can be interpreted as consumer demand profiles: a part of the cognitive structure specifying which attributes will come to mind when a certain product use is envisaged. For this reason, we have decided to incorporate those product's attributes into the model which customers consider at purchase as they directly meet their needs and expectations, and which they are able to evaluate with their knowledge.





The following attributes, associated with the total quality of an oven, considered in the model are:

1. Functional attributes:

- The energy source needed for operating the oven: electricity or gas.

- Top and bottom oven heaters: yes or no;

- Fan heating: yes or no;

- Heating by infra-red heater and fan: yes or no;

- The mode of opening the oven door: conventional or drawer-like;

- Material in oven: enamel, catalytic liners, pyrolitic liners;

- Clock: Signal clock, electronic timer;

- Accessories: modest, average, and very good;

- Energy consumption: low, average, high;

- Buttons for hob: buttons for electric hob; buttons for gas burners and electric plates; buttons for gas burners; no buttons for hob;

- The number of different hobs which the oven can be combined with.

2. The number of colours which the oven is available in;

3. The aesthetic appearance of the oven measured with: acceptable, good, excellent;

4. Price;

5. The image of producer considered at purchasing decisions.

To measure the influence of the attribute source of energy used for operating the oven on the market share two variables are needed. With the first, the influence of gas as the source of energy is measured and with the second, the influence of electricity. 'Me functionality of the oven is measured according to its functions: top and bottom oven heaters, fan heating, heating by infra-red heater and fan. To each function of the oven one variable is assigned with which the impact of the function on the market share is measured. It is namely expected that the more functions an oven possesses, the higher its quality is. Under this assumption there is no need to define a variable for the impact of functions that the oven does not possess.

Likewise the variables for measuring the impact of other attributes are determined. The attribute button for hob is used in the model since it determines the kind of hob with which the oven can be combined. It is, therefore, expected that this attribute plays a major role when purchasing the oven or cooker. All possible values, which the attribute price can take, are divided into 22 nonoverlapping intervals. To each interval one variable is assigned. With it the impact of prices, belonging to this interval, on the market share is measured.

The attributes associated with the quality of hob considered in the model are as follows:

1. Functional attributes

- The number of standard hotplates;

- The number of fast hotplates;

- The number of automatic cooking plates;

- The number of oval plates;

- The number of halogen plates;

- The number of dual circuit cooking plates;

- Sensor buttons: yes or no;

- Indicator of functions in operation: yes or no;

- Buttons on hob: yes or no;

- The number of gas burners;

- Electric ignition of gas burners: yes or no;

- Material of bob: enamel, stainless steel, glass ce ramic;

- The number of ovens which the hob can be combined with;

2. The number of colours which the hob is available in;

3. The aesthetic appearance of the hob measured with: acceptable, good and excellent.

4. Price.

5. The image of producer considered at purchasing decision.

The majority of the attributes' values mentioned before are objective ones. Only the aesthetic appearance of the product is partly subjective. By the evaluation of aesthetic appearance, the attributes like the oven light, oven door with window, signal lamps, illuminated controls, material of front and aesthetic appearance of buttons and handles are also taken into account. Therefore, the large part of the aesthetic appearance estimate consists of the product's characteristics which are objective and only a small part of the estimate is subjective.

The increasing or decreasing influence of the individual levels within the attribute is considered with the constraints (3) and (4). To create these constraints we have to find out the nature of the attribute. In the case of the attribute oven functionality, which is measured with the three functions, it was not clear in advance which of these three functions has the largest and which one the smallest impact on consumers. For this reason, it was left to the model to determine which function is the most and which one the least important.

As far as the attribute price is considered it was clear that lower prices have greater influence on purchasing decisions than the higher ones. The described impact of prices is considered in the model with constraints (4) when the values of attribute price are arranged from the lowest up to the highest.

The similar procedure is used for other attributes. Only in the case when the nature of the attribute can be determined in advance, the constraints (3) or (4) are applied.

Almost all data needed for the evaluation of the product's attributes can be obtained in catalogues where the products are displayed with the pictures and where all important technical data is added. In some cases, the inspection of the product in a shop and the conversation with sellers were needed to obtain information not available in the catalogue. The information on quantity of each product sold on the Slovenian market has been obtained from the producers if the product is manufactured in Slovenia or from the sellers who are authorized to sell imported products on our market. Generally, there were no difficulties in collecting data needed in the mathematical model. According to their market shares the products from each segment were ranked. The rank one belongs to the product with the highest market share.

To illustrate, the values of attributes collected for one of the cookers are shown in Table 4. Taking into account the values of attributes and variables expressing the significance of the attributes given in Table 4 the theoretical market share defined by (1) for this product equals to

W2+W111+W121 +Wl3l+3Wl52+2W253+2W211 + 1.05W162 + W231 + 2W32 + W41 + 40.158W506 + W61 = U I

In the same way, data for other products are collected. They form the base for creating the model (l)-(6). In Table 5 some data about segments, and the number of optimal solutions obtained by model (l)-(6) are given.

In the first segment, i.e. Cooker, 20 different products are covered in the study. The largest differences among the products from this segment are in their market shares, and in the quantity of technical progress which is built into the products. The optimal solution obtained by the model (1)-(6) for this segment in which all 20 products are considered, shows great differences between the theoretical market shares U i defined by (1) and by the actual market shares Mj for the majority of products. Theoretical market shares for products with high market shares were much lower than actual ones, but for the products with low market share many theoretical market shares were much higher than the actual ones. These differences were so large that the optimal solution can not be accepted.

The reason for large differences can be explained by the statement that high sophisticated products with a high price achieve much lower market share than the less sophisticated products with a low price on our market. The analysis of differences between the actual and theoretical market shares has helped in the decision that this segment should be broken down into three subsegments. In the first subsegment (Sl) 8 similar cookers with high market shares are assigned. These 8 products achieved the market share of 40% in 1993. The products belonging to the second subsegment (S2) achieved the market share of 6%. The products of the third subsegment (S3) are more sophisticated products with the market share of 3%.

By solving the model (l)-(6) for the first, the second and the third subsegment we obtained the optimal solutions in which the differences between the theoretical and the actual market shares are equal to zero, what is shown in the column Sum of differences in Table 5. Because the same fact has been established for the other segments too, we can conclude that the model (1)-(6) yields zero differences between the theoretical and the actual market shares if only the products bought by consumers with similar purchasing behaviour are considered in the model. It can be concluded that each subsegment presents a group of customers with similar purchasing behaviour. This information is especially useful if we consider that in commercial applications, market segmentation ranks among the primary purposes for performing conjoint analysis, both in US (Wittink and Cattin, 1989) and in Europe (Wittink et al., 1994). Benefit segments are formed on the basis of common preferences, so that products or services can be optimally designed or targeted.





For each subsegment, the model (1)46) has more than one optimal solution. All optimal solutions for each subsegment were calculated and its number is given in the column Number of optimal solutions in Table 5. Considering all optimal solutions obtained for the k-th subsegment, the average optimal value of variable wipk denoted by w*ipk is equal to


The calculation of the average optimal value w*ipk defined by (9) is made by programme QuattroPro. To illustrate, contributions to the market shares obtained by the described approach for cookers assigned to the first subsegment, are given in Table 6.

In the first column of Table 6 the set of the cooker's attributes with the positive values wi;k is given. in the next columns, the contributions of the attributes to the products' market shares, calculated in accordance with (7), are given. The market share of the first ranked product PI is 15.92. The cooker's price which is arranged to the interval 40-45 contributes the largest part to the market share of product P1. The prices are given in 1000 SIT. Taking into account the contributions of attributes to the other products' market shares from this subsegment we can conclude that the price is the most important attribute and it determines the purchasing decision in this subsegment, The importance of the price and the other cooker's attributes is presented in the last column of Table 6.

To make comparison of the consumer behaviour among the subsegments easier, the values obtained in the last column of Table 6 are normalized. These values for all three subsegments of the first segment are given in Table 7. We believe that these values reveal the customers' purchasing behaviour.

As it can be seen from Table 7 the most important attribute in the first subsegment is the price. Taking into account the size of this subsegment we can conclude that the price plays the most important part when purchasing the cooker. In the third subsegment, where more sophisticated cookers with high prices are assigned, the functions of hob play the most important role in purchasing decisions. It is interesting to note that the oven with its functions does not play a major role neither in the first nor in the third segment.



In the same way, the weights expressing the significance of attributes for products, which belong to second and the third segment, are calculated and analyzed.


Taking into account the obtained results we can conclude that the described approach largely contributes to a better understanding of the consumers' perception of the product's attributes.

The main results of this approach allow the manager to simultaneously identify benefit segments, classify the competing products into the segments, and reveal the significance of the product's attributes within each benefit segment. This information is helpful in selecting the marketing instruments, appropriate for each benefit segment. Under the assumption that the customers' preferences are not changed suddenly, the results obtained by this approach can be applied in the estimation of the success of an improved product as well as a new one.

The results of this model can also be used to improve the effectiveness of conjoint analysis. It, namely, works well when only some attributes are considered, and when the number of levels within each attribute is low. Applying the significance of the attributes and their most important values obtained by the described model, it is possible to substantially reduce the number of attributes as well as the levels within each attribute in conjoint analysis.



The growing number of markets in which the product must compete, and the fact that the collecting of data needed in the model (1)- (6) is very simple, contribute to the applicability of this approach in practice. It is especially useful for small and medium size companies to determine the product's attributes which ought to be improved and to select their target markets where there is a good match between the quality of their products -and the values the customers are seeking.


Band, W.A., 199 1. Creating Value for Customers. Designing & Implementing a Total Corporate Strategy. John Wiley & Sons, Inc.

Bastic, M., 1994. The optimization of Product Improvements, in: Proceedings of the 23rd Annual Conference of the European Marketing Academy, Volume 1, 19-32.

Green, P.E. and V. Srinivasan, 1978. Conjoint Analysis in Consumer Research: Issues and Outlook. Journal of Consumer Research, 5 (September), 103-123.

Green, P.E., 1984. Hybrid Models for Conjoint Analysis: An Expository Review. Journal of Marketing Research, 21 (May), 155-159.

Green, P.E. and V. Srinivasan, 1990. Conjoint Analysis in Marketing: New Developments With Implications for Research and Practice. Journal of Marketing, October, 3-19.

Griffin, A. and J.R. Hauser, 1993. The Voice of the Customer. Marketing Science, Vol 12. No. 1, Winter, 1-27.

Grunert, K.G., 1989. Attributes, Attribute Values and their Characteristics: A Unifying Approach and an Example involving a Complex Household Investment. Journal of Economic Psychology, 10, 229-25 1.

Horsky, D. and M.R. RAO, 1984. Estimation of attribute weights from preference comparisons. Management Science, 30, No. 7(July), 801-822.

Oral, M. and 0. Kettani, 1989. A mathematical programming model for market share prediction. International Journal of Forecasting, 5, 59-68.

Shocker, A.D. and V. Srinivasan 1979. Multiattribute Approaches for Product Concept Evaluation and Generation: Critical Review. Journal of Marketing Research, 16 (May), 159-180.

Wittink, D.R. and P. Cattin, 1989. Commercial Use of Conjoint Analysis: An Update. Journal of Marketing, 53 (July), 9196.

Wittink, D.R., M. Vriens and W. Burhenne, 1994. Commercial use of conjoint analysis in Europe, results and critical reflections. International Journal of Research in Marketing, 11,41-52.

Wright, P., 1975. Consumer Choice Strategies: Simplifying Vs. Optimizing. Journal of Marketing Research, 12 (February), 60-67.



Majda Bastic, University of Maribor


E - European Advances in Consumer Research Volume 2 | 1995

Share Proceeding

Featured papers

See More


Cultivating Collaboration and Value Cocreation in Consumption Journeys

Melissa Archpru Akaka, University of Denver
Hope Schau, University of Arizona, USA

Read More


Red Flag! The Consequences of Alerting Consumers to Fake Reviews

Jared Joseph Watson, New York University, USA
Amna Kirmani, University of Maryland, USA

Read More


Personal Budgeting: Does It Work?

Christina Kan, Texas A&M University, USA
Philip M. Fernbach, University of Colorado, USA
John Lynch, University of Colorado, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.