Conjoint analysis
From DDL Wiki
Conjoint analysis, also known as multiattribute compositional models, is a statistical technique that was introduced into the marketing research community by University of Pennsylvania, Wharton professor Paul Green in the late 1960s. Since then, conjoint analysis has become the most popular multiattribute choice model in marketing. It can be used as a trade-off measurement technique to determine consumer preferences and buying intentions, and to also predict how consumers might react to changes in a current product or new product introduction.
Contents |
Basic Process
The basic process for a conjoint analysis is as follows:
- The product features that will be analyzed are determined.
- Potential customers are shown numerous different sets of products. Each set includes several product profiles that have multiple conjoined product features.
- For each set of products, they select (by ranking, rating, or choosing) the profiles according to some overall criterion, such as preference, acceptability, or likelihood of purchase.
- Choice simulator models such as the first choice model, average probability model, and the logit model are used to estimate the utility of each product feature.
- This information is then incorporated into a new or existing product.
Data Collection
Instead of asking respondents what they prefer in a product directly, conjoint analysis presents to them potential product profiles that have specific combinations of attributes. These profiles can be evaluated in several ways, including:
- Rating (e.g. rating with a scale from 1-10)
- Ranking (e.g. rank best as 1st, 2nd, 3rd, etc.)
- Choice (e.g. four options, which do you choose?)
An example of a showing a product profile using rating is shown below: How likely are you to buy this computer?
| How likely are you to purchase this computer? (Use a scale from 1=10, with 1 being "not at all likely" and 10 being "would definitely purchase" |
|---|
| ibm |
The conjoint analysis is usually done using stated choice, which is data collected using a survey. The advantage of this method is that it is a controlled experiment. However, the product profiles in the survey are not always the same as products in the marketplace and are also not always desired. Another disadvantage is that it measures what people say and not what they do; hypothetical results are not always true in real life. An alternative to stated choice data is revealed preference data. This consists of using data that is collected from actual performance in the marketplace. The advantage of this is that the data is a real measure of how consumers act. However, there are problems with multi-collinearity - it is difficult to tell exactly why someone is choosing one product over another.
A full-factorial survey could be generated that tests every possible combination of attributes. However, in most cases this is not feasible due to the great amount of questions needed. It is also not even necessary, since balanced and orthogonal arrays can be used to reduce the number to a much smaller fractional-factorial design. This design can be generated using a program such as SAS Intstitute's SAS System.
| header 1 | header 2 | header 3 |
|---|---|---|
| Row header 1 | cell 1,1 | cell 1,2 |
| Row header 2 | cell 2,1 | cell 2,2 |
Simulation Analysis
The most common simulator models include the first choice model, the average choice (Bradley-Terry-Luce) model, and the Logit model. The First choice model identifies the product with the highest utility as the product of choice. This product is selected and receives a value of 1. Ties receive a .5 value. After the process is repeated for each respondent's utility set, the cumulative "votes" for each product are evaluated as a proportion of the votes or respondents in the sample. The Bradley-Terry-Luce model estimates choice probability in a different fashion. The choice probability for a given product is based on the utility for that product divided by the sum of all products in the simulated market. The logit model uses an assigned choice probability that is proportional to an increasing monotonic function of the alternative's utility. The choice probabilities are computed by dividing the logit value for one product by the sum for all other products in the simulation. These individual choice probabilities are averaged across respondents. In summary, while the literature shows the maximum utility (first choice model) to provide the best overall validation, choice behavior has a strong probabilistic component. We have not measured this component adequately, but instead attributed lack of validity to "noise", our inability to model information search and overload effects, and measurement error.
Example
Here is an example of a question on a survey
External links
- What is conjoint analysis? A useful conjoint software site that explains conjoint analysis in detail.
- SAS Software Software for designing full-factorial and fractional factorial surveys.
Reference
- Green, P. E., V. Rao, and J. Wind, Conjoint Analysis: Methods and Applications, 1999, Knowledge@Wharton.
- Green, P. E. and V. Srinivasan (1978), "Conjoint Analysis in Consumer Research" Issues and Outlook", Journal of Consumer Research, Vol. 5, (September), pp 103-123.
- Luce, R. D. and J. W. Tukey. "Simultaneous Conjoint Measurement: A New Type of Fundamental Measurement," Journal of Mathematical Psychology, 1 (February 1964), pp 1-27.
