David Travis has a good overview of collecting and analyzing usability testing data in How to record and analyze user research observations. This definition of an observation is particularly important:

Here’s the golden rule: an observation is a record of something you see or hear. Novice researchers often treat their opinions on what they like or dislike as observations. But your opinions aren’t observations — and neither are suggestions for new features or insights on how to fix usability issues. Don’t try to interpret the things you observe or fit things into a solution. That comes later.

The best observations come from watching real users do real tasks, and being as true as possible to the concrete details of what you see. You can either do this in the wild (with a field visit) or simulate actual use in a lab environment (with a usability test).

When making notes, just jot down observations, don’t try to solve the problem right there. The only thing I would add to David’s advice is that I try to mark important observations as I start to notice trends throughout the day, like so:

Now, if I could only figure out how to read my own handwriting, that would really help.