Most of us choose a research approach the way we choose a route home. Not by weighing options, but by taking the one we know. Someone trained in survey work reaches for a survey. Someone trained in interviewing reaches for interviews. The method arrives before the question has finished forming, and the question quietly reshapes itself to fit.
That is worth naming, because the honest version of this decision is short. Qualitative, quantitative, and mixed methods research each answer a different kind of question, and the question you actually care about usually points at one of them without much ambiguity. The difficulty is not knowing which. It is being willing to hear that the answer is the one you are less practiced at.
What quantitative work can and cannot tell you
Quantitative research answers questions of magnitude, frequency, difference, and prediction. How many. How often. Does this group differ from that one, and by how much. Does this variable predict that outcome once the obvious confounds are held still.
It does this well, and it does it at a scale no other approach reaches. A survey of 1,400 teachers can tell you that reported burnout is 0.6 standard deviations higher in the schools that adopted the new schedule, and that the difference holds after controlling for building size and years of experience. That is a real finding and a useful one.
What it cannot tell you is what burnout means to the person answering item seven. The instrument fixed that meaning in advance, when someone wrote the item. If the respondents understand the word differently than the item writer did, the number is precise and slightly wrong, and nothing inside the analysis will reveal it.
What qualitative work can and cannot tell you
Qualitative research answers questions of meaning, process, and mechanism. What is this like for the people living it. How did it unfold. Why did the thing that happened happen, in their account rather than yours.
Interview eighteen of those teachers and you learn that what they called burnout was largely about losing the unscheduled fifteen minutes between classes where they used to catch a colleague, and only secondarily about hours. That is not a variable anyone would have thought to put on the survey. It is the kind of finding that changes what a district should actually do.
What it cannot tell you is how many. Eighteen teachers cannot establish prevalence, and treating them as if they could is the most common way qualitative work gets misread, usually by the person presenting it. The right claim is about mechanism, not about spread.
The test for whether you need both
Mixed methods is not the union of two approaches. It is a design for questions that neither half can answer alone, and there is a plain test for whether you have one.
Write your research question in a single sentence. Then ask whether a competent quantitative study, done well, would answer it. Then ask the same of a competent qualitative study. If either answers it, run that study. Do the one thing thoroughly rather than two things partially.
If neither answers it, and the reason is that the question contains both a magnitude and a meaning, you have a genuine mixed methods question. "Did the schedule change raise burnout, and what is it about the change that produced it" is one question with two halves. Neither strand alone closes it.
Applying that test honestly kills a fair number of proposed mixed methods studies, which is a good outcome. Two strands cost roughly twice as much time, and a study that did not need the second one arrives late and no more convincing.
What mixed methods costs
Worth being direct about the price, because it is usually understated in the methods textbooks. You collect twice, you analyze twice, and then you do a third thing that neither tradition trained you for, which is the integration. You also write for two audiences of reviewers, each of whom will find the half outside their tradition underdeveloped.
The integration is where the cost concentrates and where most mixed methods papers thin out. Two results sections and a discussion paragraph claiming convergence is not integration, it is adjacency. The claim that the strands agree needs to be shown, case by case, with the disagreements shown too.
Where the work happens now
Both strands of a study end up on one desk, and that desk is increasingly a Mac. That matters more than it sounds. Interview recordings can be transcribed on the machine itself using Apple's on-device speech frameworks, so the audio never goes to a service and the data path in an IRB protocol stays one sentence long. The statistics run locally on the same machine. Once neither half of the evidence has to leave, keeping both halves together stops being a compromise and starts being the obvious arrangement.
ReliCheck MM Studio is built around the third task rather than the first two. Eight of its nineteen workflow steps are integration, which is a deliberate ratio. Joint displays are built inside the software, so the comparison between the burnout scores and the interview accounts becomes an artifact with evidence attached rather than a paragraph written from memory at midnight. Evidence strength ratings make you say, for each claim, how well each strand supports it. The integrated report comes out APA ready with both strands already in place.
The decision, restated
Pick quantitative when the question is about magnitude and you trust the instrument to mean what your respondents think it means. Pick qualitative when the question is about meaning or mechanism and you are prepared to make claims about how rather than how many. Pick mixed when the question genuinely contains both, and when you have the time the second strand actually costs.
What you should not do is pick the approach first and let the question settle around it. That is the version that produces a competent study answering a question nobody asked.
Four worked designs, with the numbers, are in Four Mixed Methods Research Examples. The integration artifact itself is covered in The Joint Display. MM Studio is at relicheck.com/mixed-methods-for-mac.