I feel much better about my thesis project after reading this post. You’re so right. How many is NOT important for qualitative research. What I’ve been trying to achieve with my thesis is pemeable description, rather than thick. Description that is rich, but still moves and can be moved.

Thanks for this amazing reminder.

Basically I would say it depends on what you’re trying to achieve. Generalizability? Then “how many” is probably the ONLY question you should ask. But designers rarely need to answer that need — they need to design innovative features that resonate deeply with social life. So “what does this mean?” is much more relevant. Down the road, you want generalize? Ok, sure, let’s do a survey or some form of analytics. But dammit, if I hear that question one more time BEFORE anyone knows anything about the subject at hand….why I…You know.

The simple truth is that we need both; we need the thick description (what should we design and why?) but in the real world we also need to know the scope of the opportunity (is it worth pursuing?).

More often than not we’re designing for a business master, and they need to make decisions beyond just what is going to work best for the audience/users/customers. They need to decide if said product even gets designed, and to do this they very much do need numbers. To tell them it’s not qual from some research high horse is not really good enough.

That’s why I’m a firm believer in giving them both qual and quant. It’s not always easy (or even achievable) but it gives the best results.

Don’t get me wrong, I love qual, and if I had to choose one or the other I’d go with qual…but I think we shouldn’t have to make such a choice if we’re smart about using all the items in our tool kit (including building bridges with our brethren in market research and customer intelligence).

In my market research days as a hybrid quant/qual researcher, I used to be careful in describing the samples for qualitative research as ‘valid’ (enough, of good quality, in depth) whereas quantitative samples were ‘representative’ (generalizable from, statistically-sound)