HILCA

The hierarchical individualized limit conjoint analysis (HILCA) is a variant of the conjoint analysis developed by Markus Voeth . It enables both the inclusion of a large number of features and the mapping of purchase intent decisions. Up to now, this has not enabled a variant of the conjoint analysis. In cooperation with McKinsey & Company and the Gesellschaft für Konsumforschung (GfK), the process has now been further developed into conjoint software that can be used for market research practice. In 2008, the software implementation was awarded the “Innovation Prize 2008 of German Market Research” from the professional association of German market and social researchers .

Problem

The analysis of purchase decision processes is one of the central fields of activity in marketing. The benefit construct is often used for analysis purposes. The conjoint analysis is a multivariate method whose main area of application is the measurement of preferences and benefits. However, the traditional conjoint analysis (TCA) only allows a limited number of attributes to be taken into account and can only be used to a limited extent for the analysis of purchase decisions. Essentially, the TCA has the following two weaknesses: on the one hand, the TCA only allows a small number of features and, on the other hand, it is hardly possible to predict purchase decisions with the TCA, since it only collects preference information about the possibility of not buying not allow. To eliminate these weaknesses, different variants have been developed. The following figure is intended to provide an overview of the approaches developed.

The overview of the conjoint processes that have been widespread to date (see figure) shows that practically all process variants focus only on one of the listed central points of criticism of the TCA: either a larger number of features should be included or an improvement in the ability to forecast purchase decisions should be brought about. The aim of the HILCA is both the integration of a larger number of features and the most realistic possible representation of purchase decisions.

Basic idea of the HILCA

The HILCA tries to meet the described TCA problems of the inadequate ability to forecast purchase decisions and the low number of features in one approach. An improved ability to predict purchase decisions is achieved by taking into account the idea of limit conjoint analysis. On the other hand, in order to be able to map a larger number of features within the procedure, the HILCA relies on the theoretical framework of information processing theory (IVT). This assumes that individuals undertake a hierarchization and then successive processing of the information blocks to be processed in order to avoid cognitive overload in complex assessment tasks. Based on the multi-attribute judgment problem within the conjoint analysis, this means that subjects exclude the evaluation alternatives from further analysis in a first step, the feature KO-forms and the important for them from the total number of available characteristics features extracting, based whose appraisal of the objects is made. If the number of these features exceeds the maximum number of features that can be considered in parallel in trade-offs due to limited cognitive capacity , then individuals break down the group of all important features into subgroups. Particularly important features are assessed most intensively in comparison to one another, while less important features are then examined with a graded intensity.

Process steps of the HILCA

In order to implement this basic idea, computer-aided interviews are conducted with test subjects. The following survey steps are carried out in these interviews:

Selection of relevant features
Compositional evaluation of the values of relevant features and naming of KO values
Decompositional rating of product concepts
Setting the limit (naming a rating value that separates acceptable and unacceptable stimuli)

Selection of relevant features: Here the test person receives a complete list of all features and characteristic values. He chooses the features to look out for when purchasing this product. If a characteristic is not selected, it can be assumed that it does not play a role for the test person in the decision. Therefore, it is no longer considered in the further course of the survey.

Compositional evaluation of the characteristics of relevant characteristics and naming of KO characteristics: From the group of characteristics previously classified as noteworthy, the characteristics should now be extracted to which the test person attaches particular importance and whose characteristics he will therefore probably deal more intensively with in the purchase decision process . In order to determine these (conjoint) features, the test person is presented with the features previously selected by him one after the other with their respective values. The test person now has to rate each individual expression on a point scale from 0 (expression is unacceptable, KO expression) to 100. In order to derive an initial assessment of the significance of the characteristics classified as relevant and thereby determine the characteristics on which the subsequent conjoint-analytical investigation task is based, the maximum range of point evaluations of the characteristics remaining after the exclusion of the KO characteristics is determined for each characteristic. By comparing these ranges, the most important features can be derived. It can be assumed that the characteristics with high ranges are of particular importance for the change in utility. The subsequent conjoint-analytical assessment task is therefore carried out with regard to the characteristics with the highest ranges of the preceding compositional assessment.

Decomposition rating of product concepts: In this step, as is usual in conjoint analysis, the test person is presented with product concepts for comparative rating evaluation, which are characterized by the five most important individual characteristics. Here, orthogonal main effect designs are used, which enable the mapping of all main effects with the greatest possible reduction in the number of stimuli to be assessed. Depending on the number of expressions for the individually most important characteristics, the test persons should rate between 8 and 25 stimuli on a scale of 0 to 100.

Setting the limit card: In the last step, the so-called limit card is set in order to separate acceptable from unacceptable stimuli. For this purpose, the test person sees all product concepts once again with the point ratings they have assigned in descending order. He now indicates up to which product a purchase would be possible for him. In individual cases, all or none of the products may be acceptable.

Benefit assessment at HILCA

While the utility value calculation for the particularly important features assessed conjoint-analytically in HILCA is carried out in the manner typical in the limit conjoint analysis, only compositional utility assessments are available for the other important, but not exposed, significant features. In order to make the benefit assessments of the characteristics generated by different methods comparable, e.g. B. for the purpose of subsequent market segmentation, the compositional utility values are to be expressed in the scale level of the conjoint-analytically generated utility values. For this purpose, a regression of the centered point values of individual characteristic values on the corresponding utility values is carried out for each individual test person. The regression coefficient obtained is used together with a level correction to convert the remaining point values into utility values. This is done individually for each respondent according to the equation:

{\ displaystyle U_ {Pa}}

=

{\ displaystyle P_ {za} \ cdot \ beta \ cdot {\ frac {S_ {n}} {S}} _ {c}}

With

${\ displaystyle U_ {Pa}}$ = Partial utility of the value a converted from point values
${\ displaystyle P_ {za}}$ = centered point value of the value a (point value of the value minus the arithmetic mean of all point values of the feature)
${\ displaystyle \ beta}$ = Regression coefficient between centered scores and conjoint part worth worths
${\ displaystyle S_ {n}}$ Average range of all centered point values of the features that did not go through the conjoint procedure (excluding unimportant features)
${\ displaystyle S_ {c}}$ = average range of all centered point values of the features that have run through the conjoint procedure

Subsequently, the handling of KO and mandatory values must be determined. Since mandatory values are considered important, they should be assigned a corresponding utility value. In order not to emphasize the characteristic described by the mandatory value above or below average, it is advisable to base this utility value on the average weight of the characteristics. It is therefore advisable to assign must values to half of the largest range across all characteristics. On the other hand, the KO values are treated at HILCA in such a way that their presence automatically leads to a non-purchase in subsequent market simulations.

In the above calculation of the level correction, the KO values are not included in the calculation of the centered point values. In other words, the arithmetic mean of the point values of a feature, which is subtracted from each point value of the feature for centering, is formed without the influence of the zero value of a possibly existing KO expression. Also, the span of the features and the average span formed therefrom and are based on a knockout expression without consideration of a possibly existing zero value. However, the features with a mandatory value are included in the calculation of and are given a range of 0. ${\ displaystyle S_ {n}}$ ${\ displaystyle S_ {c}}$ ${\ displaystyle S_ {n}}$

Validation of the HILCA

Comparative studies for feature-based methods show on the one hand that the HILCA is significantly superior to the ACA, which dominates market research practice in the field of conjoint analyzes with many features , with regard to predictive validity. In the literature, this is justified, among other things, by the fact that the ACA is based solely on a “technical survey trick” to deal with a large number of characteristics in interviews. On the other hand, the HILCA builds on the knowledge of information processing theory and thus has a theoretical foundation with regard to the question of how people in complex, multi-attribute decision-making situations cope with the information processing task in a targeted manner. On the other hand, empirical studies on the validity of conjoint variants, which aim to increase the number of integrable features, make it clear that the validity of these procedures as a whole can still be increased. This is currently being worked on, particularly through improvements in the area of activation and cognition modeling.

literature

Markus Voeth: 25 years of conjoint analysis research in Germany. In: Journal for Business Administration. Vol. 69, 2nd supplement, 1999, pp. 153-176.

Footnotes

↑ ^a ^b Markus Voeth: Benefit measurement in buying behavior research. The Hierarchical Individualized Limit Conjoint Analysis (HILCA) . Deutscher Universitäts-Verlag, Wiesbaden 2000, ISBN 3-8244-9035-8 .
^ Christian Hahn & Markus Voeth: Limit Cards in Conjoint Analysis. A modification of the traditional conjoint analysis (= Papers of the Institute of Economic operation systems and system technologies . No. 21 ). IAS, Münster 1997.
^ Adriane Hartmann & Henrik Sattler: Commercial Use of Conjoint Analysis in Germany, Austria and Switzerland (= Research Papers on Marketing and Retailing . Volume 6 ). University of Hamburg, 2002.
^ Jan Hendrik Kraus: Price setting in the equity fund business. An empirical analysis of the buying and price behavior of private fund investors with the help of conjoint analysis . Kovač, Hamburg 2004, ISBN 3-8300-1399-X .
↑ Markus Voeth & Maike Bornstedt: HILCA or ACA? An empirical comparison of computer-aided methods of multi-attribute measurement of utility . In: Die Betriebswirtschaft (DBW) . No. 4 , 2007, p. 381-396 .
^ Allen Newell , JC Shaw & HA Simon : Chess Playing Programs and the Problem of Complexity . In: IBM Journal of Research and Development . tape 4 , no. 2 , 1958, p. 320–335 , doi : 10.1147 / around 24.0320 ( PDF; 2.172 MB ).

[voeth2000-1] Markus Voeth: Benefit measurement in buying behavior research. The Hierarchical Individualized Limit Conjoint Analysis (HILCA) . Deutscher Universitäts-Verlag, Wiesbaden 2000, ISBN 3-8244-9035-8 .

[voeth1997-2] Christian Hahn & Markus Voeth: Limit Cards in Conjoint Analysis. A modification of the traditional conjoint analysis (= Papers of the Institute of Economic operation systems and system technologies . No. 21 ). IAS, Münster 1997.

[3] Adriane Hartmann & Henrik Sattler: Commercial Use of Conjoint Analysis in Germany, Austria and Switzerland (= Research Papers on Marketing and Retailing . Volume 6 ). University of Hamburg, 2002.

[4] Jan Hendrik Kraus: Price setting in the equity fund business. An empirical analysis of the buying and price behavior of private fund investors with the help of conjoint analysis . Kovač, Hamburg 2004, ISBN 3-8300-1399-X .

[5] Markus Voeth & Maike Bornstedt: HILCA or ACA? An empirical comparison of computer-aided methods of multi-attribute measurement of utility . In: Die Betriebswirtschaft (DBW) . No. 4 , 2007, p. 381-396 .

[6] Allen Newell , JC Shaw & HA Simon : Chess Playing Programs and the Problem of Complexity . In: IBM Journal of Research and Development . tape 4 , no. 2 , 1958, p. 320–335 , doi : 10.1147 / around 24.0320 ( PDF; 2.172 MB ).