Maxwell proposes five components — goals, conceptual framework, questions, methods and validity — that are not walked through in a line but adjusted against each other every time one of them moves. His most useful claim is that a research design exists whether or not anyone writes it down: left implicit, it is still there, with its flaws invisible.
What I took from it
The validity threats, which he reduces to two: the bias of the person researching, and reactivity, the effect of asking. Both turn up in every process interview. You listen harder to whatever confirms the solution you already pictured, and on the other side people describe the process as it should be, not as it is. Since then, before an interview I write one line on how I could be wrong and what different evidence would tell me.
The requirement that every question in the guide be justified: what it is for and which decision the answer changes. A list of topics is not a guide; it is a conversation that will drift wherever the interviewee takes it.
And memos: writing while collecting, not at the end. What you understood on Tuesday gets contaminated by what you understood on Thursday, and the order of the findings matters when you have to reconstruct why you concluded something.
What I left out
The whole academic apparatus: paradigms, the literature review, designing a thesis proposal, and the sample dissertations filling the appendices. It is also not a short book for what it argues; the practical core is two chapters, questions and validity, and the rest can be skimmed.