Ask a room of doctoral students what mixed methods research is and most of them can give you the definition. Ask what one looks like and the room goes quiet. That gap is why "mixed methods research examples" gets typed into a search box so often. The definition says a study combines quantitative and qualitative evidence. True, and close to useless when you are staring at a blank proposal, because it tells you nothing about the shape: what gets collected, in what order, and what the order is for.
The order is the design. Below are four, each with a study attached and numbers in it, because a design is easier to recognize than to define. The examples are composites, drawn from the kinds of studies that come through often enough to be familiar, not from any one project.
One. Explanatory sequential: numbers first, words explain them
A district adopts a new math curriculum across fourteen schools and wants to know whether it worked. The quantitative strand runs first. Post-test scores, 412 students, an independent samples t test against the comparison schools, d = 0.31, p = .008. A real effect, and a modest one.
That result raises a question it cannot answer. Three schools showed roughly twice the average gain. Two showed none. So the second strand goes looking for why, with sixteen teacher interviews sampled deliberately from the buildings at both ends.
The tell for this design is the sampling. The quantitative results chose who got interviewed. If you could have drawn the interview sample before seeing the numbers, you are not running an explanatory sequential study.
Two. Exploratory sequential: words first, numbers generalize them
A campus wants to measure belonging among first generation students and finds that the existing scales were written for a different population at a different kind of institution. Measuring with a borrowed instrument would produce numbers that mean something, just not the thing anyone wanted to know.
So the qualitative strand runs first. Twenty-two interviews, coded, themed, and then read again specifically for the language students used about fitting in. An eighteen-item instrument gets built from that language. The survey goes to 640 students, and an exploratory factor analysis returns three interpretable factors with alphas of .88, .84, and .79.
The tell here is that the qualitative strand produced the instrument. The words are not illustration for the numbers. They are the reason the numbers are valid.
Three. Convergent: both at once, then compared
Ninety-six nurses in one hospital system, one six-week window. The quantitative strand is a burnout inventory. The qualitative strand is twelve interviews. Neither waits for the other. Each is analyzed on its own terms, and only then are the two set beside each other in a joint display.
In this case the scale showed emotional exhaustion high across the whole unit, which a reader might have predicted. The interviews attributed it almost entirely to scheduling and short-notice shift changes, and barely at all to patient load, which most readers would not have predicted. The scale could not have surfaced that. The interviews could not have established how widespread it was.
The tell is simultaneity, and the payoff is usually disagreement. When the two strands agree you have confirmation, which is worth having. When they disagree you have the most interesting paragraph in the paper.
Four. Embedded: one strand inside a larger design
A randomized trial of a reading program, 1,100 students, is the study. Inside it sits a small qualitative strand: nine classroom observations and nine teacher debriefs, asking one narrow question, whether the program was delivered the way it was designed.
That strand is not there to produce its own findings. It is there so that a null result is interpretable. Without it, a trial that shows nothing leaves you unable to say whether the program does not work or was never really implemented. Those are different papers and different recommendations.
The tell is subordination. The embedded strand serves the main design rather than standing beside it.
Picking one
The order follows the question, not preference or training. If you have a result and need to explain it, the words come second. If you need to measure something no existing instrument captures, the words come first. If you want two independent readings of the same situation, run both and compare. If your worry is implementation, embed.
Worth saying plainly: designs in the wild are messier than four labels suggest. Studies shift mid-flight, a convergent design turns explanatory when the first results surprise everyone, and a reviewer asks for a strand nobody planned. Naming the design is a way to stay honest about what the study can claim, not a promise to the methodology gods.
The part nobody searches for
All four examples are easy to describe and hard to execute, and the difficulty is never in the t test or in the coding. It is in the join. Integration is what makes a study mixed rather than two studies stapled together, and it is the part most software leaves you to do by hand, in prose, late at night.
This is where the machine in front of you starts to matter. Both strands of a mixed methods study live on one researcher's desk, and that desk is increasingly a Mac. Interview audio can be transcribed on the machine itself with Apple's on-device speech frameworks, which means the recordings never go to a service and the data path in an IRB protocol stays one sentence long. The statistics run locally. Nothing about the work requires shipping either half of your evidence somewhere else, and once that is true, keeping both halves in one place stops being a compromise.
ReliCheck MM Studio was built around the join rather than around either strand. Eight of its nineteen workflow steps are integration, which is an unusual ratio and a deliberate one. Joint displays get built inside the software instead of assembled in a word processor, so the convergent example above produces an actual artifact: burnout scores and interview themes in one table, aligned by group, with the places they disagree visible rather than argued for. Evidence strength ratings ask you to say how well each side supports each claim. The integrated report comes out APA ready, with both strands already in it.
What changes
The practical difference is where the integration paragraph comes from. Written by hand at the end of a long project, it tends to be the thinnest part of the paper, a claim that the strands converge, made mostly from memory. Built alongside the analysis, it comes with a display behind it, a record of which evidence was weighed, and a defensible answer when a committee member asks how the two halves were joined.
That is the whole argument for taking the design seriously up front. Not rigor for its own sake. A finished paper where the most important section is the strongest one instead of the weakest.
More on the shape of the software in Mixed Methods Deserves Software Built for It, and on the integration artifact itself in The Joint Display. MM Studio is at relicheck.com/mixed-methods-for-mac.