What is Conjoint analysis?
Conjoint analysis is a marketing method allowing us to know the expectations of consumers about a product and to model their choices. The conjoint analysis method is now extremely common in marketing. Modeling of consumer choice is a key area of marketing. Conjoint analysis is used to simulate competitive markets using a single analysis.
Conjoint analysis is a method that helps you find out the expectations of consumers towards new products and to model their choices - both crucial steps of a marketing analysis. Two methods of conjoint analysis are available: full profile conjoint analysis and choice based conjoint analysis (CBC). XLSTAT is a complete statistical program which allows you to run through all the analytical steps of conjoint analysis which can be divided in five steps:
- Choice of the relevant factors and their modalities to describe the products.
- Generation of design of experiments based on full factorial, fractional factorial, D-optimal and incomplete block designs.
- Collection of the results in Microsoft Excel sheets.
- Data analysis with specific regression methods - MONANOVA (monotone regression), multinomial logit, conditional logit, hierarchical bayes etc.
- Simulation of new markets with various methods: first choice, logit, Bradley-Terry-Luce, randomized first choice.
These steps can be carried out both for a full profile conjoint analysis and for a choice based conjoint analysis (CBC).
In this tutorial, we will detail the steps necessary for the implementation and interpretation of a conjoint analysis with XLSTAT.
Dataset to conduct a conjoint analysis
An Excel spreadsheet containing the results of this example can be downloaded by clicking here. The results are divided into different sheets:
- Factors: this sheet contains the characteristics of the selected factors.
- CBC Design: this sheet contains the profiles generated, and the choices given by the 10 individuals.
- CBC: this sheet contains the results of conjoint analysis by hierarchical Bayes approach (CBC).
- Simulated market: this sheet contains the complete market to simulate.
- Market Simulation: this sheet contains the results of the market simulation.
First step: Choosing the factors
In this tutorial we will look at a classic case of conjoint analysis based on the introduction of a new product in a competitive market. This product is a drink (tea).
A brand of softdrink want to introduce a new product and is using conjoint analysis to answer two questions. Which characteristics should be found in the drink in order to, first, please the greatest number of people, and, secondly, gain market shares in an already competitive market?
The first step in the conjoint analysis is done in collaboration with experts in the beverage market. We focus on choosing the important characteristics to define the drink. The selected factors are:
- temperature (very hot, hot, iced)
- sugar (no sugar, 1 sugar, 2 sugar)
- Lemon (yes, no)
- intensity (strong, medium, light)
From these factors, you can get 54 different products. Judges will not be able to evaluate all these products. So we will use experimental designs to reduce the number of products presented to the respondents. In addition, in the choice based conjoint analysis (CBC), selections of products are presented to the individuals who will choose the one they would buy.
Second step: the selection of profiles and generation of the comparisons
Use XLSTAT-Conjoint analysis so you can select a number of definite profiles gathered in the form of comparisons and therefore allow the respondent to make choices.
Once XLSTAT is started, click on the CJT icon and choose the function Design for choice based conjoint analysis.
Once the button is clicked, the dialog box appears.
You can then enter the name of the analysis, the number of factors (four in our case), the number of profiles to classify (12), the number of comparisons (20, this number has to be greater than the number of profiles) and the number of profiles per comparison (3).
In the "Factors" tab, use the option "selecting data in the worksheet" and select the data in the "Factors" sheet. Do not select labels associated to each column.
In the Output tab, do not activate the individual sheets for this example, they are not necessary. In a comprehensive analysis, they can be very useful in order to get the results filled directly by the individuals.
Once you click the OK button, a new dialog box appears. This allows you to select the fractional factorial design of experiments or to optimize the design (D-optimal). We use the "optimize" option.
Once you click the Optimize button, the calculations are made, then the results are displayed.
The first table summarizes the generated model. The second table is the table of profiles
The following table is the table of choice, found in the "CBC" sheet and must be completed after the individuals have been interviewed. The choices are between 1 and 3 for each individual. The numbers on the left of the table are associated with profiles of the profile table.
Step 3: Fill the conjoint analysis tables
The conjoint analysis tables can either be filled directly after interviewing individuals about their choices externally or directly using the individual sheets and automatic referencing of results. This is especially interesting in the context of CBC analysis because completing the overall table can be complex.
Important: When using the optimize version of the design algorithm. Please check the validity of the design by responding to a set of question and running the model. If utilities are well displayed, then, you can make your respondents fill the rest of the table.
Step 4: Results of the analysis
As part of this analysis, 10 individuals have been questioned about their preferences in terms of tea. The results are in the CBC sheet.
To start the analysis, click the icon CJT and choose the function conjoint analysis based on the choice . You can then select the data.
Select the 10 columns of the table of responses completed by individuals as answers. Select the three columns of the numbers of choice (without the names of the selections) as a table of choices and select the profile table as profiles (without the names of the profiles).
In this tutorial, we use the hierarchical Bayesian approach. In the option tab, we select the method “Hierarchical Bayes” and we leave all the other values as default.
Once you click the OK button, the computations are performed and the results are displayed.
The most important results are the utilities and importances. They can be found in the first tables. Contrary to the classical approach (conditional logit model), the utilities and importances are given individually.
This shows that most of the individual consider that the temperature factor is an important factor in their process choice. About the utilities, we note that the lemon has a negative effect (80% of the individual). If needed, the detail for each individual can also displayed.
Step 5: Market simulation
The main advantage of conjoint analysis is to simulate a market even if the products in the market have not been tested by the individuals.
In our case, the market for a tea-based beverage is analyzed and we would like to know the impact and market shares associated to a new product. This product is a strong iced tea with lemon and no sugar. We know that in today's market there are 4 tea-based beverages that have different characteristics, the following table shows the simulated market:
To start the simulation, click the CJT icon and choose the function conjoint analysis simulation . You can then select the data.
Utilities are those obtained in the CBC sheet, the table of information about variables is the one obtained in the CBC sheet. The simulated market is in the simulated market sheet (do not select the names of products). You can also select the name of the product just behind the Product ID button. Select the CBC model and the logit method for simulation.
Once you click the OK button, the calculations are performed and the results are displayed.
The table shows that the market share for the new product are greater than 20%. This result seems satisfactory in order to launch the product on the market.