Three worked mixed methods research examples - education, health, and philanthropy - each drawn to the joint display and the integrated finding.
A mixed methods example is only worth studying at one point: the place the quantitative strand and the qualitative strand produce a finding together. Most published examples skip it — a numbers section, a themes section, and a discussion paragraph that gestures at integration. The three studies here are drawn the other way, around the joint display and the one finding neither strand could reach alone.
By Unmesh Sheth · Founder & CEO, Sopact · Updated May 2026
A mixed methods research example is a worked study that shows how one research question was answered with both a quantitative and a qualitative strand — and, the part that matters, how the two were integrated into a single finding. A useful example shows the joint display and the meta-inference, not just the two strands side by side.
An example that stops at "here is the survey, here are the interviews" teaches the data collection. The examples worth copying show the integration — the step that makes a study mixed methods rather than two studies in one document.
Most published examples are built like a binder: a quantitative chapter, a qualitative chapter, and a discussion that says the two "complement" each other. Copy that structure and you have learned to run two studies and staple them. The examples worth studying are built around the joint display — the table where the strands meet — and they end on the finding neither strand could reach alone.
The three examples below are each drawn to that one point. You will see the quantitative finding and the qualitative finding for each — but the line worth keeping is always the third one, the integrated finding, where a number and an account become a single conclusion the study can act on.
Why integration is the whole study, and why it is easiest when the strands share one record, is the case made on the pillar: mixed methods research, redefined.
One in education, one in health, one in philanthropy. Each shows the two strands, the place they meet, and the meta-inference — the conclusion the study was commissioned for. The numbers below are illustrative, drawn to show the shape of an integrated study.
91 percent of the cohort is in good academic standing at the end of year one.
In the term-two check-ins, about 30 academically-fine scholars write about isolation, money strain, and thinking of leaving.
GPA is a lagging indicator of dropout risk. The narrative names the risk two terms before the transcript does — so the study becomes an early-warning risk profile, not an end-of-year report card.
The academic dashboard reads 91 percent fine. Fourteen of those 30 scholars leave by year two — a loss the numbers never flagged.
Run wave over wave, the two strands together show not just who is at risk but when the risk first appeared.
Completion rose 14 points overall — but stayed flat at 3 of the 11 sites.
Patients there describe no evening hours — shift work and childcare make a daytime appointment impossible.
The program works. The constraint at the flat sites is scheduling, not navigation — so the fix is evening hours, not more navigators.
The 3 flat sites read as program failure. The budget response would be more navigators — money spent on the wrong lever.
The quantitative result chose who to interview. The qualitative phase existed only to explain the flat sites.
24 of 40 grantees met their targets. Sixteen missed.
Read against their reports and audits, 11 of the 16 misses share one cause: a delayed state matching grant.
The misses are not 16 isolated failures. They are one shared external risk — so the foundation's exposure is concentrated, and bridge funding, not closer monitoring, is the lever.
Sixteen grantees flagged as underperformers. Sixteen separate corrective conversations. The systemic funding risk stays invisible.
The 150-page reports, audits, and financials were read on arrival, not skimmed once at the end.
Each example follows the same shape: one question, two strands, a point where they meet, and an integrated finding. That shape is the anatomy of every worked study — the next section names its four parts.
Strip the three studies above to their structure and the same four parts appear. A mixed methods example missing any one of them is not yet an integrated study.
The study asks a single question both strands are built to answer. If the survey has its own question and the interviews have another, the example is two studies, not one.
The quantitative and qualitative strands attach to the same respondents or records. Only then can a single construct be read across both — the rating and the account, side by side.
The table that puts the quantitative finding and the qualitative finding for each construct on one row. Without it, integration is a claim in a paragraph, not a result you can check.
The finding stated at the end that neither strand could reach alone. It is what the study was commissioned for, and in a good example it is the line you remember.
Read a published study against these four parts. If you cannot point to the joint display and the meta-inference, you are looking at a data-collection example — useful for instruments, not for mixed methods data analysis.
The three studies are not the same shape. Each picks a different sequence of the two strands — and the sequence is the design. Two run convergent, with both strands collected together; one runs explanatory sequential, with the numbers chosen who to interview.
| Example | Design | What ran first | What the design made possible |
|---|---|---|---|
| Scholarship risk profile | Convergent parallel, run wave over wave | Both strands together, every term | The narrative flags dropout risk in the same term the GPA still reads fine. |
| Clinic appointment study | Explanatory sequential | The quantitative strand — appointment rates across 11 sites | The flat-site numbers chose exactly who to interview, so the qualitative phase only had to explain. |
| Grantee cohort review | Convergent parallel, merging | Both strands together, inside the same grant reports | Forty reports read against 40 metric sets in one pass surfaced the one cause behind 11 misses. |
The design decides when each strand is collected and which one leads. For the full set of designs — convergent, explanatory sequential, exploratory sequential, embedded — and how to choose one, see mixed methods research design.
The three studies above each end on an integrated finding. In a textbook, that finding arrives in a report months after the data was collected. The grantee review read 40 reports over 150 pages each; the scholarship study tracked 220 records across terms. Done by hand, the integration is the slow part — and the finding is late.
Sopact Sense holds the closed-ended score and the open-ended answer, the transcript, and the 150-page report on the same record under one persistent Contact ID. A versioned rubric reads each narrative against the measures on arrival, construct by construct. The joint display assembles as the data lands — so a study like the three above produces its meta-inference in time to act on it, not in time to file it.
Bring a study with both strands collected. We will map them onto one record, build the joint display, and read it to the integrated finding — on your data, not a sample set.
A clear example: a clinic study measures follow-up appointment rates across 11 sites (the quantitative strand) and finds rates rose at most sites but stayed flat at three. Interviews at the flat sites (the qualitative strand) reveal those clinics have no evening hours. Integrated, the two strands produce one finding the numbers alone could not: the program works, but the flat sites need a scheduling fix, not more staff.
A mixed methods study has four parts: one research question both strands answer, a quantitative and a qualitative strand attached to the same units, a joint display that puts both findings on one row, and a meta-inference stated at the end. A study missing the joint display or the meta-inference is two parallel studies in one document, not an integrated one.
A scholarship program tracks term GPA and credits (quantitative) alongside open-ended check-ins and advisor notes (qualitative) for a cohort, every term. The numbers show 91 percent in good standing; the narratives show a cluster of academically-fine students writing about isolation and thinking of leaving. The integrated finding: GPA is a lagging indicator of dropout risk, and the narrative flags the risk two terms earlier.
A patient-navigation program is evaluated by measuring follow-up appointment completion across 11 clinics, then interviewing patients at the sites where rates did not move. The quantitative strand locates the problem; the qualitative strand explains it. The integrated finding names the real constraint at the flat sites, which the appointment numbers on their own could not identify.
A foundation reviews 40 grantees by pulling closed-ended outcome metrics from grant reports and, at the same time, reading the 40 narrative reports, audits, and financials. Both strands are collected together and merged. The metrics show 16 grantees missed targets; the reports show 11 of those misses share one external cause. That comparison of the two strands for the same units is a convergent parallel design.
The clinic appointment study is an explanatory sequential design. The quantitative phase runs first and finds three of 11 sites flat. That result then shapes the qualitative phase: only patients at the three flat sites are interviewed, and the interviews exist specifically to explain the flat result. The numbers chose who to talk to.
A joint display is a table that places the quantitative finding and the qualitative finding for the same construct on one row, then states the integrated finding. For example, one row reads: numbers show wages flat for 56 of 240 trainees; narratives show most non-risers had job offers fall through; integrated finding, the flat wages are an external-market effect, not a program failure.
A meta-inference is the conclusion drawn from the integrated strands together, one neither could reach alone. In the grantee example, the numbers show 16 misses and the reports show a shared delayed grant; the meta-inference is that the misses are one concentrated funding risk, not 16 isolated failures — so bridge funding, not closer monitoring, is the response.
Write the integrated finding first, not last. Present the research question, then the joint display, then read each row for whether the strands confirm, explain, or contradict, and lead the conclusion with the meta-inference. The two per-strand analyses are supporting detail. A write-up that ends with "the findings complement each other" has skipped the integration the study was commissioned for.
A case study is a deep examination of one case and may use one method or several. A mixed methods example specifically shows quantitative and qualitative strands being integrated into a single finding. A case study can be a mixed methods example when it carries both strands and a joint display; many case studies are qualitative only and are not.
Yes. The scholarship example is both: it integrates a quantitative and a qualitative strand, and it runs the integration every term rather than once. A longitudinal mixed methods example rebuilds the joint display at each wave, so the integrated finding can be tracked over time — which turns a study into an early-warning risk profile.
Published mixed methods studies appear across education, health, and policy journals, and the Journal of Mixed Methods Research is a dedicated source. When reading any example, check it against the four parts: one question, two strands on the same units, a joint display, and a meta-inference. An example that stops at two parallel result sections will teach data collection, not integration.
This page covers the worked examples. The pillar holds the methodology and the redefinition; the other four guides cover the design types, the analysis step, the survey instrument, and the tools.
A working session, not a demo. Bring a study with both strands collected — ratings, transcripts, reports, whatever you have. We map them onto one record, build the joint display, and read it to the meta-inference. You leave with a worked example of your own study.
Live walkthrough · 30 min · with Unmesh Sheth, Founder & CEO · bring a study with both strands collected
