Conjoint Analysis - Study of trade -offs and preferences

Conjoint analysis, also called multi-attribute compositional models or stated preference analysis, is a statistical technique. Conjoint analysis requires research participants to make a series of trade-offs. Analysis of these trade-offs will reveal the relative importance of component attributes. To improve the predictive ability of this analysis, research participants should be grouped into similar segments based on objectives, values and/or other factors. It is concerned with understanding how people make choices between products or services or a combination of product and service, so that businesses can design new products or services that better meet customers’ underlying needs.

The analysis also helps us make models to predict the preferences of the customers, using the existing data. One such exercise we did in class was to give preference/rank/rate/score to the quality levels we wish to have in our spouses. Many variables were discussed which added to 600 odd options which were then reduced to 8 and finally 18 options. These options thus can be used to predict the preferences of the various participants. This technique can thus be used to predict the preferences and thus design new product or services.

Using Conjoint Analysis, the value that individuals place on any product is equivalent to the sum of the utility they derive from all the attributes making up a product. Further, it assumes that the preference for a product and the likelihood to purchase it is in proportion to the utility an individual gains from the product. There are three phases in the analysis of conjoint data: collection of trade-off data through a questionnaire, statistical analysis of the data, and market simulation.

Conjoint analysis applies a complex form of analysis of variance to a respondent’s choice task data to calculate a utility for each level of each attribute. These are basically index numbers which measure how valuable or desirable a particular feature is to the respondent. The idea is each respondent’s choice tasks reveal something about the relative utility that he or she has for each feature. Features which a respondent is reluctant to give up from one choice task to another are judged to be of higher utility to that respondent than features which are quickly given up.

A respondent’s “utility” is a measurement of his or her relative strength of preference for each level of each attribute of the research company. The respondent’s utilities are estimated using a “least squares updating” algorithm. Initial estimates of utilities are based on the respondent’s rank orders of preference and his or her ratings of attribute importance. Estimates are updated following each trade-off task, and the initial estimates have decreasing influence as the interview progresses.

The final estimates are true least squares, with the same weight being applied to each of the respondent’s answers. Utilities scaled in this way are ideal for predicting the likelihood of acceptance; they can be very misleading when reported in the aggregate or for comparing segments. For these purposes, utilities are re-scaled in such a way that the sum of the differences between the maximum and minimum level of each attribute equals the number of attributes times 100. This method assures that all survey respondents’ utilities are equally rated in reports and analyses.

The best way to interpret utilities involves analysis of the gaps between utility levels within an attribute. This “gap” or range between utility levels within an attribute indicates that the survey participants see greater importance between certain attribute levels than between other attribute levels.

Thus this analysis can be used for finding opportunities to shrink product lines, testing whether additional products cannibalize or add to preference, and uncovering segments and aligning products with their preferences.

Author:- Shweta Bhosale(13106)
Group :- Operations 3