Know-How Exchange : Research/Metrics

Hi

The last time I commented on conjoint analysis I was told off for defining it as a process involving:

Guesswork (to identify what you should analyse), Survey, Analysis and more guesswork, Stats and weighting results (More guesswork), Inventing what you want to find and then reporting it!

Of course, rigorously applied with meticulous attention the statistical relevance of each stage and a lot of care in structuring your surveying technique will minimise these problems, but few conjoint analyses actually achieve this rigour, so the result is, sadly, often normative nonsense which could have been gleaned via a focus group.

There is a very good definition on:

https://www.quickmba.com/marketing/research/conjoint/

This is disappointing because whilst it refers to the methodology, it does not reference the other sources the author uses.

A better reference is:

https://www.populus.com/techpapers/conjoint.php

This offers some extremely useful sources including a couple of PDF downloads and some excellent case studies. Where in my mind some of these fail is that they do not indicate the level of application of statistical technique that is required at each stage of the process. Whilst the first download is very interesting, it leaves the reader wondering how the criteria on the matrix were decided on in the first place and give little in the way of guidance as to whether it was done by some prior analysis or by a bunch of new age romantics dancing round a stone circle armed with pendants and crystals!

The second download is from a Sawtooth Guru and whilst well argued, read a bit like Albert Einstein giving the results of a paper and saying, “take it or leave it!”

A last resource for now is:

https://www.questionpro.com/conjoint/

You will find some interesting examples, methodology and some DIY examples. Questionpro’s raison d’etre is selling online surveys, so you would expect them to be keen to tutor their users into getting some worth from them.

Finally, (rather than lastly!) this site has quite a few references to conjoint techniques and they would be worth referencing.

Steve Alker
Unimax Solutions

I