Help with a Conjoint-ish question survey setup | XM Community
Question

Help with a Conjoint-ish question survey setup

  • 14 August 2023
  • 5 replies
  • 79 views
R
Rank
Userlevel 5
Badge +11

Hi All,

 

I have a list of discounts which we want to ask our guests for them to choose which ones tey feel would offer best value for money and then ask the question in a slightly different way - which one they would most likely prefer

 

The master list could be something like:

Free room upgrade
Free cancellation
Free night stay
Free spending money
£25 off per room
£35 off per room
£40 off per room
10% off
15% off
20% off
 
--------------------------------------
 
Q1, 3 choices from the above list chosen at random:
Which of these offers do you feel offers best value for money [single choice]
Free room upgrade
Free night stay
£40 off per room
 
==page break==
 
Q2, 3 choices from the above list chosen at random again:
Which of these offers do you feel you would most likely prefer [single choice]
Free cancellation
Free night stay
10% off
 
note that by chance ‘Free night stay’ has been chosen again, which is fine.
--------------------------------------
 
 
Some questions...
 
1) Is this technically still a conjoint question (except we’re only looking at 1 feature and not the normal minimum of 2)
2) We have access to the conjoint templates but they all seem to limit you to a minimum of 2 features.  Is there any way we can do this analysis or am I getting confused with what type of analysis I should be doing.  Perhaps this more of a preference test but then how would you analyse the grouping between the two questions?
 
Any guidance on the above would be greatly appreciated.  The starting point is to compare the different groupings of the list of discounts to understand which is more preferred over others.  The rest is flexible.
 
Thanks
 
Rod Pestell

5 replies

R
Rank
Userlevel 5
Badge +11

HI All,  A little more research and maybe what I am looking for is something called ‘monadic analysis’ as I’m wanting to focus in on a single feature.  This might also be known as tradeoff analysis but then that sounds like it will take the choices from a previous question and then ask the respondent to narrow them down further.

 

Any guidance would be appreciated.

 

Thanks

 

Rod Pestell

TomG
Rank
Userlevel 7
Badge +27

Rod,

What you want to do is really two ‘most only’ maxdiffs.  Unfortunately, a Qualtrics Maxdiff survey doesn’t support that. Could you perhaps get what you need with one most/least maxdiff?

We actually do what you want to do frequently to evaluate features, but we don’t use the Qualtrics Maxdiff tool. We do our own design and implement it as a MC question in a loop.

 

R
Rank
Userlevel 5
Badge +11

Hi Tom, 

Thanks for the reply.  It sounds like a multichoice question is the (only) way to go then but presume then there is no built in analysis setup to run?  Or... can you import the survey into a Concept Testing Program project and it will let me tell it what need analysing (haven’t read up on what type of data and layout it would expect but presume this is probably a no go as it’s likely to only look at ‘concept testing’ projects rather than standard surveys).

 

Can you shed some light on the resulting analysis process or any more details, eg. in what form or layout would we need to get the data in to process it?

 

I’ve used Loop and Merge only a little so will have a read up on it.

 

Thanks

 

Rod

R
Rank
Userlevel 5
Badge +11

Hi @TomG 

Just wanted to provide an update.

 

I’ve been working through a number of options. I setup a MC question with random order and limited to 4 sets as suggested.  The analysis though is where we’re stuck as there is no guided package.  Therefore, as a backup I’ve developed a means (in excel) which just measures which is the most commonly selected offer and also the most common one selected in a combination of options.  It’s not statistical at all but perhaps will come in handy one day.

 

So the above is in effect now my backup solution and as of this week we got some advice in that the type of conjoint we need to do is an Adaptive Conjoint Analysis.  Does this make sense?  Not heard of that before and still trying to find an example aligned to our needs to be sure that will work.

 

It’s also been suggested we outsource to a company called Conjointly.com.  Have you heard on them?  It looks like they provide the adaptive conjoint and when you buy an annual subscription they offer support.

 

Conjoint analysis is brand new to me and so I think that sounds like a good option as they will hopefully guide us through and help us understand the analysis.  And then once I’ve understood the analysis perhaps then it will be possible to go back to using Qualtrics once we know how best to analyse the data, as in the long run it would make better sense to do this all in Qualtrics (supported by excel no doubt!).

 

Thanks

Rod

TomG
Rank
Userlevel 7
Badge +27

Rod,

Based on your original description of what you are trying to do, it doesn’t make sense.  Adaptive Conjoint Analysis is for when you have too many attributes and you only have 10 features. Maybe your requirements are different from what you originally described.

Tom

Leave a Reply


Terms & Conditions