[Sigia-l] Card Sorting: How Many Users to Test

[Donna Maurer]

You know, when it comes to card sorting, I'm yet to see a domain where there are 
significant differences between segments (however, I could imagine significant 
differences if there are language or similar cultural differences, which I haven't ever 
examined). This could be due to the domains in which I work, so I'd love to hear 
specific examples.

The key differences that I have seen come from people who understand the content 
and people who don't. People who understand it sort it the 'right' way (according to 
them), people who don't understand it sort it on surface characteristics.

Donna

PS - I don't mean statistical significance - I've also never bothered to measure this ;) 
Insight is more important than correlation.


On 19 Jul 2004 at 13:01, Jonathan Baker-Bates wrote:

> > > You must test fifteen users to reach a correlation of 0.90, 
> 
> I'm unclear how these correlations square with the "who to test" issue.
> Most clients have desired segments they want represented in their
> testing (Teens, Housewives, Bald Women... OK maybe not that divergent,
> but sub-sections of their customer base), so is he talking about the
> number of users you have in the *same* segment here? If in a test group
> of 20, the Asian Doctors group things radically differently from the
> Truck Drivers and Tea Ladies, surely the fact that there are only five
> Asian Doctors is significant.
> 
> Jonathan
> 
-- 
Donna Maurer
blog: http://maadmob.net/donna/blog/
work: http://steptwo.com.au/
AOL IM: maadmob