Conjoint analysis

Conjoint analysis ( Jointly Considered - " holistically "), even composite measurement, conjoint measurement, conjoint analysis, multi -attribute compositional models, trade-off analysis is a multivariate method that was developed in psychology. It goes next to the measurement of the evaluation of a (possibly fictitious) material in the focus of the dekompositionellen method to measuring the importance of an individual component ( variable ) of the total benefit. These specific characteristics of the goods ( stimuli) are provided with specific importance weights, in order to derive a possible universal overall preference judgment the consumer of the good can.

• 6.1 The limit conjoint analysis (LCA )
• 6.2 The Hierarchical Individualized limit conjoint analysis ( HILCA )
• 6.3 The Multi -Rule- conjoint analysis (MRC )
• 6.4 Choice Based Conjoint analysis with hierarchical Bayes estimation ( CBCHB )

History

The in the works of Luce and John W. Tukey (1964 ), first mentioned in conjoint analysis was introduced in the 70s by Paul E. Green and V. Seenu Srinivasan as a method in market research and analysis is the method most commonly used today for collecting the consumers' preferences. Using conjoint analysis examines the extent to which individual features or combinations of features that distinguish a particular product, are preferred by the user.

Example

For a car manufacturer, it would be important to determine the significance of the features of " manufacturer ", " power " and " car color " for the purchase decision of the user. In a conjoint analysis, a number of total products would be combined from these features (eg, a red Audi with 170 hp, gray Mercedes with 160 hp and a blue BMW with 190 hp, etc.). The respondent is now to these overall concepts from each one vote. As part of the conjoint method, it is possible to conclude from the information of the user on whose preferences regarding the individual characteristics and characteristic values ​​. In our example, this could occur, for example, that the subjects buying a new car manufacturers are focusing primarily on ( main feature ), the manufacturer BMW would be preferred ( most important feature expression ).

Since each created by a company Good can be interpreted as a combination of product characteristics with certain characteristic values ​​, the method Conjoint has found widespread that it owns in the market and marketing research today.

Process principle

Essential for the understanding of conjoint analysis is the so-called decompositional principle of this method: The ratings of respondents initially relate to holistic product combinations to those features and their characteristics are broken down in the evaluation and converted, which have been included in the evaluation. This procedure corresponds to the conjoint analysis to a large extent the actual evaluation process of a real purchase situation in which the consumer is also faced with holistic products. Since these products have respondents from the perspective of both certain advantages and certain disadvantages, it is brought to weigh relative to each other the meaning of the various properties and to get the actual meaning of the individual characteristics aware.

Areas of application of conjoint analysis

The three most important areas of application of conjoint analysis in marketing research can apply the areas of product development, pricing and market segmentation. In the area of ​​product development conjoint analysis play a major role, especially in the launch (launch ) of new products and the re-launch of existing products and to be modified. " What are the characteristics of my product are the ones who cause the maximum benefit to the buyer? " Could, for example, a question in this context, be that would be clarified in the implementation of a conjoint analysis in the corresponding sample. Results of such surveys on new products, in addition to the increase in sales also help to save costs, as conjoint analysis may those product characteristics as irrelevant for the buyer identified, which are connected to produce a high cost.

In the area of ​​pricing conjoint analyzes are often used to provide the data base for the calculation of the estimated price-demand function for a product in a given market or in a competitive environment. With the data of conjoint analysis can be carried out a so-called market simulation over which can calculate one price for a given product to the manufacturer brings the profit optimum. In addition, it can be predicted with market simulations, as competitors respond to market launches or find strategies is best to react to such innovations from competitors.

Two standard methods of conjoint analysis

The basic form of conjoint analysis has been transferred over the years in numerous variants, which are intended to overcome certain weaknesses of the traditional process. There are two main disadvantages should be mentioned: the one hand, in the original version, the number of features that can be queried very limited. On the other hand provide rating and ranking questions, as used there, no direct conclusion on the actual product selection of respondents, which are the basis of a market simulation. Among the conjoint methods that have evolved over the years due to modifications and specialization of existing methods, two methods have enforced that allow a better management of those problems: the Adaptive Conjoint Analysis (ACA ) and the Choice Based Conjoint Analysis ( CBC).

The adaptive conjoint analysis

In contrast to the "classical" conjoint methods, the Adaptive Conjoint analysis is a method that is just computerized feasible. Therefore, this method is referred to as adaptive because the entries of the subject are already processed during the interview by the computer and used to develop the respective next questionnaire page. The interview thus adapts to the individual preference structure of the individual user to draw meaningful information as possible from the interviews. For the respondents, the Adaptive conjoint analysis presents itself as a rather varied nature of the survey, as an ACA consists of five phases questioning that he has to go through. Here, the computer learns the preference structure of the subject from one phase to know you better and designed the questionnaire pages each so that they get the maximum information value.

In contrast to classical conjoint analysis, the ACA is not a full -profile method, that is, that the subject in the course of the interview never has to evaluate product combinations that are composed of ALL features. Each of the products to be valued rather consists of only a small number of features, with information about the preference structure of the subjects arise in the course of the interview yet with respect to all features. Although increases with a higher number of features, the length of the interview, because the subjects more preference judgments are required. However, the interview lengths vary even with large research designs in moderate magnitudes. In practice, ACA studies are usually with 8-15 features and each about five characteristics performed theoretically there is a possibility of up to 30 features in the survey design incorporated. Since it is difficult for the respondent to give consistent ratings for a complex situation, the effect of important features can be at ACA observe that characteristics that are " actually " for the subject unimportant overestimated tend to differ from him, however, be underestimated tend. This is particularly problematic when pricing or market simulations are a key objective of the study.

The choice -based conjoint analysis ( choice-based conjoint analysis)

The second advance in the field of conjoint method represents the choice -based conjoint analysis, called the CBC. In contrast to traditional conjoint analysis is at the CBC to a method which is based on the findings of economic decision theory. By mapping the decision situation of consumers mainly two improvements are achieved: First, the forecasts are based on the accurate and reflect reality, secondly, they are behavioral and thus sales oriented directly. The purchase probabilities obtained in the course of choice-based conjoint analysis can be converted directly into expected gross margins, earnings and market shares.

In contrast to the ACA is in the CBC to a full -profile method, ie, the subject rated products, as composed in real purchasing situation, always from all possible features. Another difference from the ACA is that the user can not grade its own assessment of the submitted products. The user gets rather per page survey a number of products submitted, from which he only select one as the preferred by him, ie " purchase " can.

Since the subject must weigh at such a situation, for example, four products, each with all its features against each other, the CBC provides much higher demands on the attention of the study participant as an ACA. In turn, can be determined more accurately from the thus obtained responses, the trade- off between the individual features. There are also implicit decision criteria obvious that the respondents are not necessarily aware of.

For longer interviews, however, a learning process can take place at the respondents, in which he no longer perceives the product in its entirety, but only due to him less relevant features (eg, brand, price ) decides. The task mimics the buying process, in which the buyer hides unimportant features and focuses on relevant criteria. As long as the information reduction corresponds to that of the real purchasing process, which is often observed, this is not very problematic.

Currently, choice-based conjoint analysis of the "gold standard " of the industry (in Sawtooth Solution Newsletter (2006) was published that 75 % of their customers CBC, 16 % ACA and 9% use the Traditional conjoint method. ). An interesting study on the CBC provide Albers / Becker / Clement / Papies / Schneider: " measurement of willingness to pay and their use for price bundling " in Marketing ZFP, 29 Jg (2007), p.7 - 23rd An answer to the question why conjoint analysis, and in particular CBC works, Joel Huber supplies.

Developments

The limit conjoint analysis (LCA )

The limit conjoint analysis (LCA ) extends the traditional conjoint approaches to a further process step. The subjects will be presented a certain number of stimuli in the first step which must be evaluated in accordance with the survey design and put into a ranking. A stimulus here is a combination of different property forms.

In the second step, the individual purchasing Willingness now be obtained using the stimuli are divided into kaufenswerte and non- kaufenswerte alternatives. This is done by placing a limit Card ( LC) after the last still kaufenswerten rank space. Here, the LC can not be set exclusively between two stimuli, but also before the first or after the last rank position. How can express the subject that he respect any or all stimuli having a willingness to buy.

The LC is interpreted as zero benefit. Kaufenswerte take positive stimuli, not kaufenswerte become negative utility values ​​. This approach makes the assumption necessary for the subjects to assess the benefit distances between the ranks as a constant. Also the groups as " worth buying " and " not worth buying " classified stimuli must be assumed to be equal scaled. Be In this way, in contrast to the classical conjoint analysis, absolute utility values ​​, rather than determined by mere changes in utility.

This is a key weakness of traditional conjoint analysis - namely, that the prognosis of purchase decisions is hardly possible, since only preference information is collected, but do not allow the mapping of non- purchases - eliminated, but also allows the LCA only recording a small number of features.

The Hierarchical Individualized limit conjoint analysis ( HILCA )

The HILCA only a more improved forecasting ability of purchasing decisions by considering the idea of ​​limit conjoint analysis. Moreover, to be able to represent a larger number feature within the process, reaches the HILCA back on cognitive theories. These assume that individuals make a hierarchy, followed by successively processing the processed information to avoid cognitive overload in complex assessment tasks. Through a hierarchical assessment approach for features, which performs a different kind of hierarchy in comparison to the hierarchical conjoint analysis, and the basic idea of ​​limit conjoint analysis can be the HILCA mark that allows both the inclusion of a theoretically unlimited number of features and the purchase decision intended to improve prognosis.

The multi -Rule- conjoint analysis (MRC )

The multi -Rule- conjoint method (MRC ) into account several (multi) psychological decision rules ( rule ) of the respondents. This may reflect not only rational but also irrational decisions this method as opposed to the traditional conjoint analysis. Since the irrational as well as rational decision-making behavior has a specific system, it can be statistically calculate and predict.

The rational decision-makers in the traditional conjoint models of computation, weighted the characteristics of a good individually and shall render its decision by summing the part -worths to an overall benefit. An irrational decision is based, however, on certain benchmarks, such as price reductions, and comparing the different offers directly from these properties. For him it is important to decide what alternative for most properties with respect to the respective reference value is the better option and not know which rationally in the sum has actually a greater overall benefit.

Through a combination of statistical algorithms for the prediction of rational and irrational decision-making behavior, the prediction quality can be significantly increased.

Choice Based Conjoint analysis with hierarchical Bayes estimation ( CBCHB )

Using a hierarchical Bayesian (HB ) models the preferences of individual persons Estimate within a data set. It should be noted that the hierarchical Bayesian estimation is a special statistical estimation procedure and should not be confused with the hierarchical structure of an assessment task. Where individually too little information on the assessment of individual features is present, it is derived in the framework of the HB estimate of the preferential distribution of the total population, which leads to very robust results. The distribution of preferences in the population is of particular interest, since through them the heterogeneity of the customer is mapped. From the heterogeneity of the customer population can eg be derived, what percentage of customers ever sufficiently pronounced preferences has to give them to offer the product profitable. Furthermore, the hierarchical Bayesian method avoids distortions in the application in the simulation of profits and profit margins, which inevitably occur in the aggregate method by Jensen's inequality, as soon as the population distribution in general has a heterogeneity. A new variant of the method is the Adaptive Choice -based Conjoint Analysis ( ACBC ).

Example

One group of participants is a product ( MP3 player) presents which four characteristics (maturity, capacity, features and warranty ) in each of two values ​​( high, low ) has.

In the following, the volunteers, the product will be introduced, each with different versions of the four characteristics. Examples of these configurations would be

• High capacity and run time, but levels of equipment and warranty
• High capacity and features, but low runtime and warranty
• Guarantee high and features, but low runtime and capacity

The participants should now indicate their order of preference, by ordering the differently configured goods in the order that corresponds to their preferences.