The second lowest price has a better purchase likelihood than the lowest price

Question

Hi,

I have a CBC Analysis containing 208 examples. Within the CBC I have six price levels. The CBC-HB utilities for the price attribute are:
9 €             66.37
19 €      64.78
29 €       37.81
39 €      -12.26
49 €     -49.12
59 €     -107.57

Now I wanted to use SMRT to create a Generic Sensitivity Analysis (no specific assumed competition) to make the report a bit "nicer" and clearer. ( I used your description: http://www.sawtoothsoftware.com/about-us/news-and-events/sawtooth-solutions/ss27-cb/1109-reporting-preferences-for-attribute-levels-in-conjoint-analysis).  

But after I ran the simulation 19 € has a higher purchase likelihood than 9 €.  
9 €            91,53
19 €    95,27
29 €    89,95
39 €    57,14
49 €    30,3
59 €    13,56

Do you have any idea why the higher price now has a better purchase likelihood than the lower price? Is it because the two CBC-HB utilities are so close to each other? Maybe I also chose the wrong "average base", as I am actually not sure what average score for non-ordered attributes (like brand or color) exactly means?


Thank you very much for your help!

Answer 1

You might find the following two articles helpful that discuss potential sources of these types of issues:

Respondents prefer higher prices to lower prices: http://www.sawtoothsoftware.com/support/knowledge-base/design-and-methodology-issues/900-price-utilities-showing-higher-price-has-more-utility

Utilities and simulations are not in agreement: http://www.sawtoothsoftware.com/support/knowledge-base/general-issues/1456-average-utilities-and-simulations-are-not-in-agreement

Comments

Thanks for answering, but could it also be that i chose the wrong base at SMRT, because I am not sure what is meant with average score for no-ordered attributes like Brands?

---

I believe the article is just saying to use the brand that is closest to the average of all the brands.  Off the top of my head I don't think that should be problematic because the issue is in the ordered attribute of price, and a higher price produces a higher preference.  You will probably find that to be the case even if you switch to each of the brands in your study.  I imagine it's probably one of the issues in those articles dealing with reversals and non-normally distributed utilities causing the averages to be a little misleading.