[Sigia-l] research: when is it enough? vs. ethics

[PeterV]

That was interesting!

One of the things I was trying to find out was how can you plan your 
research and know it will be enough - too much. For example, I once planned 
and did a test with ten people. I actually only needed about 7, the three 
last ones only confirmed (valuable in itself I know) what I had already 
learned. If you are a beginning IA, and planning your first large research 
exercise, how do you know how much to plan for?

On a related note, I remember Bruce Tog talking on a conference about the 
"3.58 correlation" (maximum correlation is always one): he would throw 
numbers like that at management to convince them, *even* if the numbers 
really didn't mean anything by themselves. I've done the same: you have 
done the research, reached some good (qualitative) conclusions, but to 
convince the client you spend an hour making some (qualitative) 
calculations (that maybe aren't really statistically relevant) and mention 
some good looking numbers in the report.

I realise this is be on the verge of being non-ethical (at least it's 
non-academic), and don't do this often, but you could also look at it as 
communication skills. Or sales. A similar thing happens when quoting 
studies. We all know if we quote some Forrester study and throw in some 
numbers, it makes you look more convincing for many business people. Even 
though you personally may have doubts about the relevance of that 
particular study. As long as you believe in the core point you are making I 
think that can be valid practice in certain circumstances. After all, it 
will benefit the client in the end.  Then again, I wouldn't want to present 
a report that *relies* on numbers I don't believe in. Just uses them as a 
little extra convincing power for the CEO or project manager not used to 
qualitative research. We can be as academic as we like, this is what 
happens in the real life trenches of IA. It works.

I realise this may be controversial, any opinions?
PeterV
http://petervandijck.net