Different research methods can be used at different stages
Interviews done with actual users are the mainstay of user research, as far as I’m concerned. Such an interview (also sometimes known as an IDI or “individual depth interview”) should ideally be conducted in the participants’ normal context, or environment in which they normally undertake the activity you’re studying. You can have a set of questions to guide you—called “semi-structured” interviews—or you could keep it completely open—called “unstructured”.
Tip: another potentially useful variation is known as the proxy interview (also known as ‘key informant’ or ‘intermediary’ interviews).
Small group discussions around one or more topics, often used in (and given a bad name by) market research but usually with a different purpose in mind and using a different approach. They are good for idea generation, brainstorming, comparing alternative designs. The social interaction can elicit more information than if you interviewed each person individually. However, they’re not so good because you can’t see how people do what they do, only what they say they do. Managing the group can be difficult and peer pressure can affect the answers given (known as “group think”).
Tip: storytelling sessions (aka “anecdote circles”) are a good alternative to a traditional focus group. These techniques are gaining popularity in management and business collaboration fields.
Workshops are a good framework for a variety of smaller research methods that can be strung together to make the most of your time with participants. Almost any “workshop activity” can be used, depending on what it is you’re trying to achieve; I’ve listed two below, but you can find more at design games
Tip: most of my “focus groups” are actually workshops, but the former is a term better understood by stakeholders.
Whilst normally used as an IA design technique, card sorting is an excellent way to research an audience’s mental models and terminology, particularly as part of a workshop. For instance, what do they call XYZ? where is there disagreement on the classification of content/things? what is the organisation’s mental model? (when running a stakeholder workshop). In a group situation, pay attention to the discussion and debate, not just the obvious output (the “quantitative taxonomic data”).
Tip: regarding that taxonomic data, you can use Donna Spencer’s card sort analysis spreadsheet to help see interesting patterns.
A contextual inquiry is essentially an unstructured interview in the context in which users use the system. The researcher acts as the student and learns everything they can from the user (eg how they perform their job, what tools they use). As with a lot of direct contact research methods, it’s not so much about what you find out about a particular product/system/website/application, but more about getting access to people you know are your audience. You want to learn about what makes them tick, who they are beyond a one-dimensional demographic breakdown—which is so often how the audience is seen by organisations.