Mixed methods research is an approach that deliberately combines qualitative and quantitative data within a single study or programme of inquiry, integrating the two strands to answer questions that neither could fully address alone. As Creswell and Plano Clark define it, mixed methods research involves collecting, analysing and “mixing” both forms of data, using the combination to produce a more complete and corroborated understanding of the research problem than a single approach would yield.
You should reach for mixed methods when your research question has both a measurable, generalisable component (how much, how many, how strongly related) and an interpretive, meaning-rich component (why, how, in what context). The defining feature is not simply using two methods side by side, but integrating them so the whole becomes greater than the sum of its parts.
What is mixed methods research?
Mixed methods research is the third major methodological movement, sitting alongside purely quantitative and purely qualitative traditions. It rests on a pragmatic philosophy: rather than asking which paradigm is “true”, the pragmatic researcher asks which combination of methods best answers the question at hand. Where a quantitative study counts, measures and tests relationships, and a qualitative study explores meaning, experience and context, a mixed methods study does both — and, crucially, brings the findings into conversation with one another.
The intellectual case for mixing rests on the idea of complementary strengths. Quantitative data offer breadth, precision and the ability to generalise from a sample to a population; they answer “what” and “how much” with statistical confidence. Qualitative data offer depth, nuance and the ability to surface unanticipated explanations; they answer “why” and “how” in the participants’ own terms. By design, the weaknesses of one strand are offset by the strengths of the other. If you are still deciding which tradition fits your project, our guide to quantitative vs qualitative research is a useful starting point before you commit to mixing.
It is worth being precise about what mixed methods is not. Running a survey and also reading a few documents is not automatically mixed methods. The label only properly applies when both strands are conducted rigorously, when there is an explicit rationale for combining them, and when the two sets of findings are integrated rather than simply reported in separate chapters. Integration is the hallmark of the approach.
“Mixed methods research is an approach to inquiry involving collecting both quantitative and qualitative data, integrating the two forms of data, and using distinct designs that may involve philosophical assumptions and theoretical frameworks.” (Source: Creswell & Creswell, 2018)
When and why to use mixed methods
Greene, Caracelli and Graham’s influential framework sets out four classic purposes for combining methods. Map your research question onto one of these before you choose a design — the purpose drives everything that follows.
- Triangulation — you use qualitative and quantitative data to converge on, corroborate or cross-validate findings about the same phenomenon. Agreement across methods strengthens confidence; divergence prompts deeper investigation.
- Complementarity — you use one method to elaborate, enhance or clarify the results of the other, capturing different but overlapping facets of a complex issue.
- Development — you use the results of one method to inform or build the other — for example, qualitative interviews that generate the items for a later survey instrument.
- Expansion — you use different methods for different components of the study to extend its breadth and range, answering distinct sub-questions with the method best suited to each.
A fifth purpose, initiation, is sometimes added: deliberately seeking paradox and contradiction between strands to provoke fresh interpretation. If your dissertation question contains both a measurable relationship and an unanswered “why”, mixed methods is likely to serve you well. If it is purely descriptive of one type, a single approach is cleaner and more defensible.
Notation: how to read QUAN, quan, QUAL, qual
Mixed methods researchers use a shorthand introduced by Janice Morse to communicate two things at a glance: the priority of each strand and the timing of the strands relative to one another.
- Capitals (QUAN, QUAL) mark the dominant or priority strand that carries the main analytic weight.
- Lower case (quan, qual) marks the secondary, supporting strand.
- A plus sign (+) means the strands run concurrently, e.g. QUAN + QUAL.
- An arrow (→) means the strands run sequentially, e.g. QUAN → qual.
So “QUAN → qual” tells a reader, instantly, that this is a quantitatively driven study in which a smaller qualitative follow-up is conducted afterwards to explain the numbers. This notation appears throughout the design descriptions below.
The core mixed methods designs
Most mixed methods dissertations adopt, or adapt, one of four core designs. The first three — convergent parallel and the two sequential designs — are illustrated in Figure 1 above; the embedded design is a variation in which one strand is nested inside a larger study of the other type.
1. Convergent parallel design (QUAN + QUAL)
In the convergent parallel (or concurrent triangulation) design, you collect quantitative and qualitative data at roughly the same time, analyse each strand separately using its own appropriate techniques, and then merge the two sets of results to compare and corroborate them. Both strands typically carry equal weight.
- When to use it: when you want to validate one type of finding against the other, or obtain a fuller, side-by-side picture of a single phenomenon in a single phase.
2. Explanatory sequential design (QUAN → qual)
Here the study runs in two phases. You begin with a quantitative phase, analyse the results, and then design a qualitative follow-up specifically to explain, contextualise or unpack the quantitative findings — particularly any surprising, ambiguous or statistically significant results. The quantitative strand has priority and drives the study.
- When to use it: when your numbers raise questions they cannot answer themselves — you have a relationship or a difference, but you need participants’ voices to interpret why it exists.
3. Exploratory sequential design (QUAL → quan)
This design reverses the order. You begin with a qualitative phase to explore a poorly understood phenomenon, use those findings to develop an instrument, framework or hypotheses, and then test or generalise them quantitatively in a second phase with a larger sample. The qualitative strand has priority in shaping the study, but the quantitative phase extends its reach.
- When to use it: when no suitable measure exists for your context, so you must first build one from the ground up — or when you want to test whether qualitative insights hold across a wider population.
4. Embedded (nested) design
In the embedded design, one strand is nested within a larger study driven by the other. The supplemental strand plays a supporting role and addresses a different sub-question. A common form is a small qualitative component embedded inside an experiment or trial — for instance, interviewing a subset of participants to understand how they experienced an intervention whose effect is being measured quantitatively.
- When to use it: when one method genuinely dominates the study but a secondary question — about process, acceptability or mechanism — needs the other method to answer it.
Mixed methods designs at a glance
The table below summarises the four core designs, their notation, the typical purpose they serve, and where integration happens.
| Design | Notation | Timing | Main purpose | Point of integration |
|---|---|---|---|---|
| Convergent parallel | QUAN + QUAL | Concurrent | Triangulation / corroboration | At analysis (merge & compare results) |
| Explanatory sequential | QUAN → qual | Sequential | Explain the numbers (complementarity) | Phase 2 built on Phase 1 results |
| Exploratory sequential | QUAL → quan | Sequential | Build & test an instrument (development) | Phase 1 findings feed Phase 2 design |
| Embedded (nested) | QUAN(qual) or QUAL(quan) | Concurrent or sequential | Answer a secondary sub-question (expansion) | Supplemental strand informs the primary |
Integration and joint displays
Integration is what separates true mixed methods from a study that merely happens to use two methods. It is the deliberate point at which the strands are brought together so that the combined insight exceeds either strand alone. Integration can happen at three levels:
- Design level: the strands are connected through the chosen design (e.g. one phase builds the next).
- Methods level: through merging, connecting, building or embedding datasets.
- Interpretation/reporting level: in a narrative that weaves the findings together, or in a joint display.
A joint display is a table or figure that arrays quantitative results alongside the corresponding qualitative findings so a reader can see, in one place, how they relate. A typical joint display has a row per theme or construct, with one column for the statistical result, one for illustrative quotes, and a final “meta-inference” column stating whether the two strands confirm, expand or contradict one another. Where they contradict, that tension is itself a finding worth reporting.
Sampling in mixed methods
Sampling is one of the trickiest — and most examined — aspects of mixed methods, because the two strands usually demand different logics. Quantitative strands favour probability sampling and larger numbers to support generalisation; qualitative strands favour purposive sampling and smaller numbers chosen for richness. Decide, for each strand, both the sampling strategy and how the two samples relate.
- Identical sample: the same participants take part in both strands.
- Nested sample: the qualitative participants are a subset of the quantitative sample — common in explanatory sequential designs, where you purposively select interviewees from survey respondents.
- Parallel sample: different participants are drawn from the same population for each strand.
- Multilevel sample: strands sample different units or levels (e.g. pupils quantitatively, head teachers qualitatively).
A frequent mistake is to apply quantitative sampling logic to the qualitative strand — worrying that 15 interviews are “too few”. They are not; qualitative samples are judged on information power and saturation, not statistical representativeness. Choosing the right approach for each strand also affects how you plan your methods of data collection.
Locating mixed methods within your wider design
A mixed methods choice does not float free of the rest of your methodology — it sits inside a chain of decisions running from philosophy down to technique. Most mixed methods researchers adopt a pragmatist stance, which licenses using whatever methods best answer the question rather than committing to a single paradigm. Working outward from your underlying research philosophy and through the layers of the research onion keeps your methodological choices coherent, so that your design, strategy and data-collection techniques all follow logically from your stated assumptions and aims.
How to design a mixed methods study: step by step
- State a question that needs both strands. Confirm your question genuinely has a measurable component and a meaning-rich component; if not, a single method is cleaner.
- Articulate your rationale. Name the purpose — triangulation, complementarity, development or expansion — and justify why mixing adds value.
- Choose the design and notation. Decide priority (which strand dominates) and timing (concurrent or sequential), then label it, e.g. QUAN → qual.
- Plan each strand rigorously. Specify the sample, instruments and analysis for the quantitative and qualitative components separately, to full disciplinary standard.
- Decide your sampling relationship. Identical, nested, parallel or multilevel — and how participants flow from one strand to the next.
- Plan the point of integration in advance. Specify where and how the strands meet (merging, connecting, building or embedding) and whether you will use a joint display.
- Collect and analyse. Execute the strands in the planned order, keeping audit trails for both.
- Integrate and draw meta-inferences. Bring the findings together, state whether they confirm, expand or contradict, and report what the combination reveals.
Worked example: one question, end to end
The box below runs a single research question through a complete explanatory sequential design, from question to meta-inference, including the arithmetic of the quantitative phase, so you can see how the pieces connect in practice.
Research question: Does participation in a peer-mentoring scheme improve first-year undergraduate retention, and if so, why?
Rationale: The “whether” is measurable (complementarity + explanation); the “why” needs students’ own accounts. Quantitative strand has priority.
Phase 1 — QUAN. A cohort of 400 first-years is tracked; 200 join peer mentoring, 200 do not. Retention to second year is measured.
• Mentored group retained: 174 ÷ 200 = 87.0%
• Non-mentored group retained: 150 ÷ 200 = 75.0%
• Difference in retention = 87.0% − 75.0% = 12.0 percentage points
• Drop-out rates: mentored 13.0%, non-mentored 25.0%
• Relative risk of dropping out = 13.0% ÷ 25.0% = 0.52 (mentored students roughly half as likely to leave)
• A chi-square test on the 2×2 table returns p < 0.01 — the difference is statistically significant, not chance.
Connecting the phases. From the mentored group, purposively select interviewees: 8 who stayed and 4 who left (a nested sample), to explain the mechanism behind the 12-point gap.
Phase 2 — qual. Twelve semi-structured interviews are thematically analysed. Three themes explain retention: (1) early sense of belonging, (2) practical “insider” knowledge from mentors, and (3) a low-stakes person to ask “silly” questions. The four leavers report that mentor contact fizzled out after week three.
Integration / meta-inference. A joint display pairs the 12-point retention gain with the belonging and knowledge themes; the strands confirm and expand each other. The leavers’ accounts add an actionable refinement: sustained, scheduled mentor contact — not just initial pairing — drives the effect. A conclusion the numbers alone could not give.
Strengths and challenges
Used well, mixed methods is powerful; used carelessly, it doubles the workload without doubling the insight. Weigh the trade-offs honestly in your methodology chapter.
Strengths
- Offsets the weaknesses of each strand with the strengths of the other.
- Produces more robust, corroborated conclusions through triangulation.
- Explains the “why” behind statistical relationships, not just the “what”.
- Builds context-specific instruments grounded in real participant language.
- Appeals to diverse audiences and supports stronger, more practical recommendations.
Challenges and limitations
- Demanding: requires competence in both quantitative and qualitative methods.
- Time- and resource-intensive — effectively two studies and an integration.
- Integration is genuinely hard; many “mixed” studies never truly mix.
- Reconciling contradictory findings between strands can be uncomfortable.
- Tight word counts make it tempting to under-report one strand.
Common mistakes to avoid
- Calling a study mixed methods with no integration. Two separate chapters that never meet is parallel reporting, not mixed methods.
- No explicit rationale. Failing to name why you mixed (triangulation, development and so on) reads as method-collecting.
- Mismatched sampling logic. Judging a qualitative sample by statistical representativeness, or vice versa.
- Imbalanced rigour. A strong survey bolted onto three thin interviews — or the reverse — weakens the whole.
- Deciding integration too late. Plan the point of integration before collecting data, not after.
- Ignoring divergence. When strands disagree, that tension is a finding — not an error to bury.
Designing a mixed methods dissertation?
Our expert academics can help you choose the right design, align your sampling and integrate both strands into one coherent methodology chapter.
Conclusion
Mixed methods research is not a way of hedging your bets; it is a deliberate, rigorous strategy for answering questions that have both a measurable and an interpretive dimension. Choose it because your question needs both strands, name your purpose, pick the design and notation that fit, sample each strand on its own terms, and — above all — plan how and where you will integrate. Do that, and you produce findings that are at once generalisable and richly explained: the distinctive payoff that only a well-built mixed methods study can deliver.
