Practical evaluation guide

What are mixed methods? Combining quantitative and qualitative data in evaluation

An explanation of mixed methods, quantitative and qualitative roles, integration steps, practical designs and cautions.

Interpret quantitative and qualitative findings together

Mixed methods deliberately combines quantitative and qualitative data to answer one evaluation question from multiple perspectives. Quantitative research identifies extent and trends; interviews and observation explore reasons, processes and exceptions; both are then integrated for interpretation. [1]

An 80% satisfaction figure alone does not show who was dissatisfied or why. Conversely, a few striking accounts do not show how widespread that experience is.

Mixed methods does not simply place numbers and accounts side by side. It first decides which data each question requires and integrates agreement and disagreement for judgment.

Read quantitative and qualitative evidence together

Choose the roles and combination of methods, then examine where findings agree or disagree.

Roles of quantitative and qualitative data

Quantitative data suits showing how many people experienced change, its extent and differences by characteristics or time. Common sources include questionnaires, use records, tests and administrative statistics.

Qualitative data explores meaning, change processes, context and unexpected effects. A narrated causal explanation is not automatically established. Quantitative data also does not represent a population merely by being numerical.

Roles of quantitative and qualitative data
What to learn	Main data used	Example
Extent of change	Quantitative data	Proportion whose behavior changed
Reasons for change	Qualitative data	Triggers and barriers to changing behavior
Differences by characteristics	Quantitative data	Retention by participation route
Unexpected effects	Qualitative data	Specific open responses or interview examples
Overall judgment	Integrate both	Improvement issues indicated by numbers and accounts

[1]

Three practical combinations

Concurrent designs collect surveys and interviews around the same time and compare results. Explanatory sequential designs first establish numerical trends, then explore reasons through interviews. Exploratory sequential designs identify issues through interviews first, then examine their extent in surveys.

Select according to initial knowledge, time, budget and target-group burden. Build integration into design rather than conveniently connecting results afterward.

These are not the only designs; qualitative work may also be embedded in experiments. Prioritize explaining why both are needed for the question.

Concurrent: collect numbers and accounts together and compare
Explanatory sequential: explain numerical trends with later accounts
Exploratory sequential: identify issues in accounts and examine them numerically later

[1]

From design to integration

First decide whether quantitative or qualitative data are needed for each question. Then design whether to use the same participants or complementary accounts from different people and when to integrate results.

After analysis, place results in one table. For example, put survey retention on the left, interview reasons and barriers on the right, and interpretation and improvements drawing on both in the center.

Separate quantitative and qualitative aspects of questions
Decide targets, timing and methods for each dataset
Decide where results will be integrated
Analyze quantitative and qualitative data separately
Compare agreement, disagreement and complementarity
Connect integrated interpretation to recommendations and limitations

Practical framework or suggestion

Disagreement is a finding, not failure

Surveys may show high satisfaction while interviews describe difficulties. Neither must be wrong: minority experiences hidden in averages, hard-to-express dissatisfaction or differences across use stages may explain it.

Do not erase disagreement. Check differences in participants, question meaning, study timing and data limitations. Then make more specific judgments about which improvements which users need.

Even with agreement, check dependence on the same participants or response tendencies
Disagreement identifies issues for further checking
Do not infer the whole picture from one dataset
Report integration processes and limitations
Check burden and personal-information protection in both components

[2]

Fictional example; not actual results

Integrate retention and users' experiences

A fictional example for illustration: compare records with voluntary interviews to improve continued use of consultation services.

Quantitative observation	Fictional qualitative observation	Integrated judgment
Lower retention among weekday daytime users	Accounts of difficulty coordinating with work	Check time constraints alongside content
Satisfaction is high	Grateful to the consultant, but procedures are not progressing	Judge experience separately from behavioral outcomes
Some users continue	A colleague's referral enabled schedule adjustment	Investigate this as a condition supporting use

What this example supports

Check whether numerical and interview targets cover the same scope. Retain disagreements as hypotheses requiring further checking.

Common errors in judgment

Publish survey and interview results separately and stop there

How to revise: Place both results and integrated interpretations together for each question

Reason: This shows how both results inform decisions.

Assume agreement makes the conclusion certain

How to revise: Check shared bias and contrary cases

Reason: The same people's self-reports may simply have been repeated.

Put it into practice next

Decide integration timing and comparison methods for each question in the mixed methods worksheet

View an example and open the worksheet Read the method explanation

Frequently asked questions

Are mixed methods only for large studies?

No. Small programs can combine brief surveys with a few interviews according to purpose and burden.

What if quantitative and qualitative results differ?

Rather than discarding one, check targets, timing, question meanings and response difficulties. Disagreement itself may be an important finding.

Do two studies automatically constitute mixed methods?

Not necessarily. Both datasets must connect to the same evaluation question and be designed for integrated interpretation.

References and checked sections

Best Practices for Mixed Methods Research in the Health Sciences
John W. Creswell / Ann Carroll Klassen / Vicki L. Plano Clark / Katherine Clegg Smith（NIH OBSSR） · 2011
Section consulted
The Nature and Design of Mixed Methods Research (main text, pp.4–9)
How it relates to the text
Intentional integration of quantitative and qualitative research; concurrent and sequential designs
Source checked: 2026-09-13 (includes AI cross-checking)
Quality in qualitative evaluation (QQE)
HM Treasury / Evaluation Task Force
Section consulted
4. The framework（Sample 7–8 / Analysis 12 / Reporting 14 / Ethics 17）
How it relates to the text
Selection rationale, different perspectives and negative cases, paths to interpretation, consent and confidentiality
Source checked: 2026-09-13 (includes AI cross-checking)

Content updated: 2026-09-13Source checked: 2026-09-13 (includes AI cross-checking)No documented human reviewIncludes fictional examplesEditorial policy and scope of checking Cite this page Report a correction

How to cite this page

EvalCompass Editorial Team. “What are mixed methods? Combining quantitative and qualitative data in evaluation”. EvalCompass, updated 2026-09-13. https://evalcompass.jun-nakatani.com/en/guides/mixed-methods/

Add the date you accessed the page. Text created by this site is available under CC BY 4.0. Quotations, external materials, and bundled fonts have their own terms of use. When reusing content, credit the source, link to the license, and indicate any changes.

Suggest a correction or improvement to this page

Change history

2026-09-13 — Distinguished original methodological sources from practical guidance and added specific fictional examples, judgment examples and next actions. Revised categorical claims about classification and measurement conditions.