Home / Summaries / Marketing Models / customer-regression-data

New products and conjoint analysis - Conjoint analysis - Aggregation

Q: How can you examine conjoint results for more than one customer?

By combining the data before regression to averages for each perception ( e.g. in this case for all 16 different ice cream variants ).

Q: What is the downside to combining data before regression? How is it solved?

That you assume all customers are the same . Solved by looking at the data and making clusters and then performing regression on the clusters.

Q: What is an alternative method to analyze conjoint results for more than one respondent?

By computing a regression per customer . Build a dataset of N rows, 1 per customer, and the new columns would be the beta's. Then look at what other information you have for each customer and look at if e.g beta 1 is correlated with gender or age. Beta's can be clustered as well.

3 important questions on New products and conjoint analysis - Conjoint analysis - Aggregation

How can you examine conjoint results for more than one customer?

By combining the data before regression to averages for each perception (e.g. in this case for all 16 different ice cream variants).

What is the downside to combining data before regression? How is it solved?

That you assume all customers are the same. Solved by looking at the data and making clusters and then performing regression on the clusters.

What is an alternative method to analyze conjoint results for more than one respondent?

By computing a regression per customer. Build a dataset of N rows, 1 per customer, and the new columns would be the beta's. Then look at what other information you have for each customer and look at if e.g beta 1 is correlated with gender or age.
Beta's can be clustered as well.