Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Conjoint analysis method to use when comparing complex products where attributes consist on availability of applications

Hello, I am helping two students using Sawtooth Software for the first time.

Basically, they wish to analyse the prefference for a wrist band that can measure dufferent things while one runs (for example heart-rate, speed, number of steps, etc...)
There is a total of 10 attributes, each consisting in the availability or absence of one of these apps (except for the Price attribute). They wish the choice question to be between different products offering a bundle of applications, for example

Band A:
Measures Heart Rate
Measures Speed
Measures Number of Steps
Priced $250


Band B:
Measures Sleep Patterns
Measures Heart Rate
Measures Lost Callories
Priced $200

Unfortunatelly for those students, I am not sure which of the conjoing analysis methods would most closelly mimic what they want. CBC would in my opinion display too much information at once, and ACBC inevitably leads to an ideal product with all attributes.
From what I understand, MBC might be closer to what I need but I am not experienced enough with that method of analysis.
asked Apr 18, 2016 by Mathias (125 points)

1 Answer

+2 votes
I agree that a design with 10 binary (on/off) attributes with price is a challenging one to do.  The problem with typical main-effects conjoint analysis (either CBC or ACBC) is that during the analysis phase, products offering mostly all "on" features at a high price will be overpredicted.  Products offering mostly all "off" features will be underpredicted.  This has been discussed at recent conferences and one way to be on the lookout for such potential problems is to be willing to collapse (during the analysis) three or four binary attributes (the most synergistic sets) into 8-level and 16-level single attributes.    It's easiest to handle this issue in CBC after data have been collected, though power users can also figure out how to recode their data during analysis for ACBC to do this as well.

I don't understand your concern about ACBC "inevitably" leading to an ideal product with all the attributes.  If you use the "Summed Pricing" approach in ACBC where if during the BYO (configurator) phase respondents have to incur extra cost to add each "on" feature, then it doesn't inevitably "lead" to the product that has all the features "on".
answered Apr 18, 2016 by Bryan Orme Platinum Sawtooth Software, Inc. (132,190 points)
Thank you, Mr Orme. Your comment was very helpful. I believe I will suggest they use the ACBC method, the tricky part for them now would be to estimate a reasonable price for having each application. I apreciate you taking your time to awnser

Although I have one last (more technical) question:
I have a subscription to the Academic Research License. When in the "ACBC Exercice Settings" dialogue, in the "Attributes" Tab, I am unable to select "Attributes" from the "List of Attributes" (in fact the List of Attributes is empty). Is it because I am using the Academic Research Licence or is this an actual problem with the program itself?

Thank you,
Mathias Praz
Actually, forget that last question, I had not read the tutorial properly
Ah, glad you figured it out!  One must first set up the attribute list and list of levels for each attribute in the "List Manager" prior to getting to this step.  Thanks for reviewing the tutorial documentation and figuring this out on your own.