"Unsupervised learning / clustering in marketing?"
Right now I have been given a clients database. The client doesn't know what to do with it, we just wants 'something' out of it.
Now I'm a very 'square' guy, I think mostly in practical terms, find how X predicts Y, for example. If I have only X to work with, I think clustering, but when I do so, I think, what for?
So I can make for instance, 3 clusters. So the client can see, OK, this group of people have high A, medium B and low C, this other group had this, and so on. I hope somehow that turns out to be useful for general info and thats pretty much it?.. as you can see I'm pretty square.
And I'm not even talking about evaluation difficulties. How many clusters? Why? How do I know they are actually statistically correct and do mean something?
Hoping to receive some insight, thank you!