Give that test a rest
The principles of qualitative social/market research are similar to those of user testing, except for one major gap:
In qualitative research, it is accepted that the results will never be fully generalisable across the population. Hence the thinking is that having more respondents will yield a rich diversity of results, provided you have the bandwidth for all that fieldwork and analysis.
In user testing (those I’ve been involved in, by the way, somewhat resemble in-depth interviews), there is supposedly quantitative proof that the optimum number of respondents per target group is five.
Five is the magic number | Credit: useit.com
I’m not sufficiently well-versed in statistics to question how that was derived, but it sounds logical. Because in user testing, you’re limiting the scope of discussion to the website (or other material) you’re asking for opinions on. So after a certain number of respondents, chances are you’ll find them sounding like a broken rccord.
Oddly though, none of the (very well-established) research houses I spoke with questioned the large number of test subjects requested. Maybe because user testing is merely a sideline for them.
Or because earnings scale with sample size.
Actually, if there’s no need for sub-group analysis, we often recommend sample size of approx. 5-6 for a typical, not-too-complex, qualitative study…. it’s usually the clients that freaked out with the small sample size. Hee. The curve looks dodgy though.
It’s from http://www.useit.com/alertbox/20000319.html. Based on the formula:
N(1-(1-L)n)
It lost me at N