Hi,
I'm trying to set up my first conjoint analysis and I have a few questions:
1. I have a list of attributes and features. However, I don't want every attribute to display in a bundle option (e.g. 10 attributes with each 3 features, I would only want 3 randomly chosen attributes in a bundle). Is this possible? When I try, it pulls in every attribute but with just one feature each. The goal is to not overwhelm the participants with so many things to consider. I would want something like this: With 10 attributes, "Do you prefer option 1 (more base pay, less vacation, same PTO) more or do you like option 2 (same base pay, more bonuses, and more flex time)."
2. In the conjoint analysis, the only option I see is forced choice between the bundles/options/packages (e.g. do you prefer package 1 or package 2). Is there way to make it a 5 point scale saying how much they prefer the options over each other?
3. Is there a way to pull in the options they voted on into a follow-up question after the conjoint exercise to evaluate each option? For example, it could be something like this: "You preferred the package with more base pay, less vacation, same PTO. If you were to receive this package, how satisfied would you be with the Company (likert scale 1-5)?" I assume some time loop and merge is needed.
4. Is this more suited for a maxdiff but just list out all features possible as attributes?
Userlevel 7
+11
Hi there! Lucky you to get to do a Q conjoint! I'm crossing my fingers to get one of these in the budget for this fall.
Here are my perspectives on your questions:
1. This is working as intended because if you don't present all the options, you won't know what caused the preference: the absence of an option or the strength of feeling for those options that were presented. Best practices are to use a maximum of 8 features, so you may want to whittle those down.
2. If you do a 5-point scale, you're not going to be able to use the tool to spit out the math correctly, as it's set up specifically for this methodology. If you're a stats wiz, you may want to instead come up with your own study and do a Key Driver Analysis instead.
3. This seems like you might be getting away from the intention of the Conjoint, which is to figure out what features and levels drive preference most, and starting to get into satisfaction. If it were me, I would do the conjoint to determine preference and then perhaps do a secondary study that delves into satisfaction, where the benefits package would be just one aspect.
Here are my perspectives on your questions:
1. This is working as intended because if you don't present all the options, you won't know what caused the preference: the absence of an option or the strength of feeling for those options that were presented. Best practices are to use a maximum of 8 features, so you may want to whittle those down.
2. If you do a 5-point scale, you're not going to be able to use the tool to spit out the math correctly, as it's set up specifically for this methodology. If you're a stats wiz, you may want to instead come up with your own study and do a Key Driver Analysis instead.
3. This seems like you might be getting away from the intention of the Conjoint, which is to figure out what features and levels drive preference most, and starting to get into satisfaction. If it were me, I would do the conjoint to determine preference and then perhaps do a secondary study that delves into satisfaction, where the benefits package would be just one aspect.
Thanks for reply,
I guess I should go into a little more detail of what I'm looking for. To address questions 1 and 2, many choice based conjoint analyses run into survey fatigue due to the number of choices with large bundles. However, it is required to have many attributes within the study to the complexity of the study (e.g. 10 attributes). I've seen in a few study even they had similar amount of attributes, they were able to choose subsets of attributes (only show 3 attributes out of 10 at a given time). For example with cars, you make consider Make, Model, Year, Color, Gas Mileage, etc.
!
It appears what I am looking for is adaptive conjoint analysis. Descriptions can be found in the link: https://www.sawtoothsoftware.com/help/lighthouse-studio/manual/hid_web_whataca.html
As for question 3, I was looking to see if there was an expansion of the dual choice feature which is a yes/no question.
Let me know if this makes sense!
I guess I should go into a little more detail of what I'm looking for. To address questions 1 and 2, many choice based conjoint analyses run into survey fatigue due to the number of choices with large bundles. However, it is required to have many attributes within the study to the complexity of the study (e.g. 10 attributes). I've seen in a few study even they had similar amount of attributes, they were able to choose subsets of attributes (only show 3 attributes out of 10 at a given time). For example with cars, you make consider Make, Model, Year, Color, Gas Mileage, etc.
!
It appears what I am looking for is adaptive conjoint analysis. Descriptions can be found in the link: https://www.sawtoothsoftware.com/help/lighthouse-studio/manual/hid_web_whataca.html
As for question 3, I was looking to see if there was an expansion of the dual choice feature which is a yes/no question.
Let me know if this makes sense!
Leave a Reply
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.