Skip to content
Back to search results

Mixed Methods Research — Integrating Quantitative and Qualitative Evidence

Type
Research design framework
Font size

What does it do?

Mixed methods research is a systematic research design in which quantitative and qualitative data are collected and analysed within a single study.

Formalised by Creswell and Plano Clark (2003, 3rd ed. 2018), it offers three principal design variants — concurrent, sequential, and transformative — and strengthens findings through triangulation across both data strands.

The intuition behind the method

Numbers alone can tell you how much or how often; words alone can tell you why or how. Mixed methods research combines both in a principled way so that the weaknesses of one data type are offset by the strengths of the other. Instead of choosing between a survey and an interview study, you design both, plan how they will talk to each other, and let the combined evidence answer questions that neither strand could address on its own.

History

Mixed methods as a formal methodology emerged in the social and health sciences during the 1980s and 1990s as researchers challenged the paradigm wars between quantitative and qualitative traditions. John W. Creswell and Vicki L. Plano Clark's foundational text, first published in 2003 and now in its third edition (2018), systematised the design typology and integration strategies that define the field. The approach draws on philosophical pragmatism — associated with Dewey and James — which holds that research questions, rather than epistemological allegiances, should drive method selection. Mixed methods is now widely used in health, education, social, and behavioural research.

When to use it

Mixed methods research is appropriate when a single data type — quantitative or qualitative alone — is insufficient to address the research question in full. It suits exploratory purposes (where qualitative findings must precede instrument development), explanatory purposes (where statistical results need contextual elaboration), comparative purposes (where outcomes must be examined across groups and contexts), and descriptive purposes that require both breadth and depth. Suitable variable types span continuous, categorical, textual, and ordinal data. Both cross-sectional and longitudinal structures are supported. A prerequisite is that separate sampling and analysis plans are feasible for each strand, and that a pragmatist stance — accepting multiple epistemological frameworks — is adopted.

Limitations and considerations

  • Demands competence in both quantitative and qualitative traditions, which is rare in a single researcher and often requires a team.
  • Requires separate, sometimes substantial, sampling and data-collection efforts for each strand, raising cost and time.
  • The integration point must be planned explicitly; poorly planned integration produces two parallel studies rather than a genuine mixed-methods design.
  • Paradigmatic tensions between positivist and interpretivist epistemologies must be consciously navigated; pragmatism is the standard resolution but not universally accepted.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Combines the breadth and generalisability of quantitative evidence with the depth and contextual richness of qualitative evidence.
  • Triangulation across strands increases confidence in findings and can reveal inconsistencies that prompt deeper investigation.
  • Flexible design types (concurrent, sequential, transformative) let the researcher match the integration strategy to the specific research question.
  • Addresses a wider range of research purposes — exploration, explanation, comparison, and description — within a single coherent study.

Applications

  • Health sciences: combining clinical trial outcomes with patient-experience interviews to explain adherence and dropout.
  • Education research: using survey data to identify achievement patterns and follow-up interviews to understand underlying causes.
  • Social and policy research: measuring the prevalence of a phenomenon quantitatively, then exploring lived experiences qualitatively to inform intervention design.
  • Organisational studies: integrating employee survey results with focus-group narratives to analyse workplace culture and change.

How it works

How it works

1

The researcher selects one of the three core design types. In a concurrent design both quantitative and qualitative strands are collected at the same time and integrated during interpretation. In a sequential design one strand informs the other — either quantitative results guide purposive qualitative sampling (explanatory sequential) or qualitative findings shape the instrument for a subsequent survey (exploratory sequential). A transformative design wraps either of the above inside an emancipatory or theory-driven framework. At this stage the integration point — whether merging happens at design, analysis, or interpretation — is explicitly planned.

2

Separate sampling and data-collection plans are executed for each strand. The quantitative strand uses structured instruments (surveys, tests, records) with a sample large enough for the intended statistical analysis. The qualitative strand uses purposive sampling alongside interviews, observations, or documents, with a size driven by theoretical saturation rather than statistical power. Minimum combined sample guidance is n ≥ 20, but each strand must individually meet its own adequacy criteria.

3

Each strand is analysed independently using the methods appropriate to it — statistical techniques for the quantitative data and thematic, content, or interpretive analysis for the qualitative data. This independence preserves the integrity of each tradition and prevents premature integration from distorting either set of findings.

4

The results of both strands are brought together at the pre-planned integration point. Common integration strategies include merging (presenting results side by side), connecting (using one strand's output as input for the other), and embedding (nesting one strand inside a larger design). Matrices, joint displays, and narrative weaving are standard tools for making the integration explicit and auditable.

5

The researcher interprets what the integrated evidence means as a whole — noting where the strands converge (triangulation), where they diverge (requiring explanation), and what the combination reveals that neither strand could alone. Pragmatist epistemology, which accepts that different methods can answer different aspects of the same research question, underpins this step.

Common pitfalls

  • Treating the two strands as entirely separate studies and reporting them independently — true integration at design, analysis, or interpretation is the defining feature of mixed methods.
  • Failing to pre-specify the integration point: decide before data collection whether the strands will be merged, connected, or embedded, and document this in the methods section.
  • Applying quantitative sampling logic (e.g., random sampling for power) to the qualitative strand — qualitative sampling is purposive and driven by theoretical saturation, not statistical representativeness.
  • Using concurrent design when one strand genuinely depends on the other's results — sequential design is the appropriate choice in that situation.
  • Ignoring divergent findings: when the strands disagree, that disagreement is itself a finding that requires explanation, not suppression.

Frequently asked questions

How is mixed methods different from simply using both a survey and an interview?

The defining feature of mixed methods is planned integration — the two strands are designed to talk to each other at a specified point (design, analysis, or interpretation). Running a survey and an interview in the same project without integrating their findings is multi-method research, not mixed methods. Integration can take the form of merging results, using one strand's output to inform the other, or embedding one strand inside the other.

Which design type should I use — concurrent, sequential, or transformative?

Choose concurrent when you want to collect both data types simultaneously and compare or merge them during interpretation. Choose sequential-explanatory when quantitative results need qualitative elaboration, or sequential-exploratory when qualitative insights must precede instrument development. Choose a transformative design when an emancipatory theory (e.g., feminist, critical race) frames the entire study and both strands serve that theoretical purpose.

How large does each sample need to be?

Each strand must meet its own adequacy criteria independently. The quantitative strand needs a sample sized for the planned statistical test (power analysis). The qualitative strand uses purposive sampling sized to theoretical saturation — typically 6–30 participants depending on homogeneity and method, not a fixed number. A combined minimum of n ≥ 20 across both strands is a rough floor, not a target.

What does it mean to adopt a pragmatist stance?

Pragmatism holds that research questions, not epistemological allegiances, should determine method selection. In practice this means accepting that quantitative and qualitative data can both contribute valid knowledge about the same phenomenon, that truth claims are evaluated by their practical consequences, and that the researcher need not choose between positivism and interpretivism as fixed philosophical commitments.

Related methods

Sources

  1. Creswell, J.W. & Plano Clark, V.L. (2018). Designing and Conducting Mixed Methods Research (3rd ed.). Sage.

    ISBN: 978-1483344379
    Copied.