What users say will generally be different from what they do -- this is true, but what's a good strategy when you can't get to enough of your users? What if you could answer some really big questions by performing simple research on ALL of your users? This is the same style of approach taken by companies like Google and Zynga, to target user research efforts which have made them what they are today. Log analysis, done well, can seem like mind-reading. If you haven't done it before: there's nothing to fear.