Everything we do needs to start with understanding our customers...or users or members or whatever you like to call those people whom you serve. Far too often though this understanding is based on anecdotal evidence -- "we hear all the time at the conference that our members hate our website" or sometimes just based on a quick look at the data, e.g. "website visits are way up though so obviously they don't hate it that much."
It's not just about collecting the right types of data (and yes, that should be plural; you'll want to access a variety of data sources to get a complete picture of your user). It's most important to start with a clearly defined goal for collecting the data in the first place.
What are you trying to learn about your user? Do you want to understand what they think of something (attitudinal) or do you want to assess how they behave under a given set of circumstances (behavioral)? Because we all know, but often forget, that what we do is often very different than what we say. So, thoughts/perceptions do not always = behavior.
From there, you can figure out what type of data will help best answer your question/fulfill your goal -- do you need how many/how much data (quantitative) or do you need to know why and how to fix (qualitative).
For example, if I want to know how people think information on my website should be organized and what we should call various content sections, I could carry out a card sort (which, if set up correctly, could give me both qualitative and quantitative data). If I'm more concerned with gathering information on how a user would behave on my website -- i.e. can they find our flagship publication and easily purchase it? -- I would choose a usability methodology.
Understanding your user doesn't have to be complicated! Start with your end state in mind and use this handy-dandy infographic to quickly see which research methodology would get you the data you need.
We’re on social media!