Quick answer: Choose quantitative when your research question asks “how many”, “how often”, “is X associated with Y” or seeks generalisable patterns. Choose qualitative when it asks “how” or “why” something happens, “what does X mean to participants”, or explores under-studied phenomena. Choose mixed methods when one method alone cannot fully answer the question — typically the strongest choice at master’s and PhD level.
By the numbers
- 52% of UK master’s dissertations use qualitative methods, 31% quantitative, 17% mixed (HESA, 2024).
- 49% of US doctoral dissertations use mixed methods, up from 28% in 2014 (Proquest Dissertation Survey, 2024).
- 12 to 20 interviews — typical sample size for thematic saturation in qualitative research (Guest et al., 2006).
- n = 384 — sample size needed for ±5% margin of error at 95% confidence in a population over 100,000 (Cochran formula).
- 0.80 statistical power threshold expected for quantitative dissertations at PhD level.
- 3 to 6 months typical analysis time for 20 qualitative interviews vs 4 to 8 weeks for an equivalent survey dataset.
Qualitative vs quantitative — at a glance
| Feature | Qualitative | Quantitative |
|---|---|---|
| Question type | How? Why? What does X mean? | How many? How often? Is X associated with Y? |
| Data form | Words, images, observations | Numbers, scales, counts |
| Sample size | 5 to 30 typically | 100 to thousands |
| Goal | Depth, meaning, mechanism | Generalisability, prediction |
| Methods | Interview, focus group, ethnography, observation | Survey, experiment, secondary data analysis |
| Analysis | Thematic analysis, IPA, grounded theory, discourse | Descriptive stats, inferential tests, regression, SEM |
| Software | NVivo, Atlas.ti, MAXQDA, Dedoose | SPSS, STATA, R, Python, Excel |
| Validity criteria | Trustworthiness (credibility, transferability, confirmability) | Internal/external validity, reliability, construct validity |
The six-question decision framework
Answer all six honestly, then count where you score Q (qualitative) vs N (quantitative):
- What does your research question begin with? “How” or “Why” → Q. “How many”, “Does X cause Y” → N.
- Has this phenomenon been studied much? Under-studied → Q (you need to map terrain). Well-studied with clear constructs → N.
- What is your primary goal? Understand meaning → Q. Test a hypothesis or estimate effect size → N.
- Can you access a large sample? No (≤30 accessible) → Q. Yes (n > 100 realistic) → N.
- How much time do you have? 6 months or less → quantitative survey or qualitative interviews (small sample). PhD timeframe → either or mixed.
- What does your discipline expect? Health/economics/psych → often N. Anthropology/education/sociology → often Q. Business/management → mixed is increasingly preferred.
If your data is textual but you still want to measure patterns, content analysis offers a middle path — the deciding factor is whether frequency answers your research question, or whether the meaning behind the words matters more.
Choosing methodology? Get a 30-minute methodology consult
PhD-qualified methodologists. Help with framing your question and choosing the right approach.
Worked examples by discipline
Business — when each fits
| Research question | Method | Why |
|---|---|---|
| “Does flexible working policy adoption correlate with employee retention rates in UK SMEs?” | Quant | Variables measurable, large sample available |
| “How do middle managers experience the transition to AI-augmented decision-making?” | Qual | Lived experience, under-studied phenomenon |
| “What drives consumer trust in influencer-marketed sustainable brands, and how does it translate to purchase intention?” | Mixed | Need both effect size (quant) and mechanism (qual) |
Nursing and health
| Research question | Method |
|---|---|
| “Does a 30-minute telephone follow-up reduce hospital readmissions in COPD patients?” | Quant (RCT) |
| “How do oncology nurses experience moral distress when delivering bad news?” | Qual (IPA) |
| “What are the barriers and facilitators to implementing nurse-led hypertension clinics, and how do they affect blood pressure outcomes?” | Mixed |
Five main qualitative analysis approaches
| Approach | Best for | Sample |
|---|---|---|
| Thematic Analysis (Braun & Clarke) | Most flexible; broad qual questions | 12 to 30 |
| IPA (Smith) | Deep lived experience of one phenomenon | 3 to 8 |
| Grounded theory (Charmaz) | Building new theory from data | 20 to 30+ |
| Discourse analysis | Language, power, identity construction | Variable |
| Ethnography | Cultural practices in their setting | Field-time based |
Quantitative test selection cheat sheet
| Question | Variables | Test |
|---|---|---|
| Difference between 2 groups | Continuous DV, binary IV | Independent t-test |
| Difference between 3+ groups | Continuous DV, categorical IV | One-way ANOVA |
| Association between two variables | Both continuous | Pearson correlation |
| Predicting an outcome | Continuous DV, multiple IVs | Multiple regression |
| Predicting binary outcome | Binary DV, multiple IVs | Logistic regression |
| Frequency comparison | Both categorical | Chi-square |
| Mediation | IV → M → DV | PROCESS macro / Baron-Kenny |
Mixed methods designs (when one is not enough)
- Convergent parallel — collect both simultaneously, compare for triangulation.
- Explanatory sequential — quant first, qual to explain unexpected patterns.
- Exploratory sequential — qual first to surface variables, quant to test them.
- Embedded — qual within quant or vice versa.
Methodology stuck? Get expert help
PhD-qualified methodologists. From research question to analysis plan.
References
- Creswell, J. W. and Creswell, J. D. (2022) Research Design: Qualitative, Quantitative, and Mixed Methods Approaches. 6th edn. Thousand Oaks, CA: Sage.
- Braun, V. and Clarke, V. (2022) Thematic Analysis: A Practical Guide. London: Sage.
- Smith, J. A. and Osborn, M. (2022) “Interpretative phenomenological analysis”, in Smith, J.A. (ed.) Qualitative Psychology. 4th edn. London: Sage.
- Charmaz, K. (2014) Constructing Grounded Theory. 2nd edn. London: Sage.
- Field, A. (2024) Discovering Statistics Using IBM SPSS Statistics. 6th edn. London: Sage.
- Hair, J. F. et al. (2022) Multivariate Data Analysis. 8th edn. Andover: Cengage.
- Guest, G., Bunce, A. and Johnson, L. (2006) “How many interviews are enough?”, Field Methods, 18(1), pp. 59–82. https://doi.org/10.1177/1525822X05279903
- Higher Education Statistics Agency (2024) UK Postgraduate Research Statistics. Cheltenham: HESA.
Need methodology help that defends at viva?
PhD-qualified researchers. Question framing, design choice, sample-size justification — all viva-ready.
Frequently asked questions
Yes — typically 30 to 50% more workload because you do both data collections, both analyses, and an integration step. But it produces a stronger contribution and is increasingly expected at PhD level in business, education and health.
Possibly, depending on effect size and design. Run a power calculation in G*Power before committing. For most behavioural studies, n < 100 makes finding statistically significant effects difficult unless effects are very large.
Neither. Both answer different question types. Strong dissertations match method to question; weak dissertations choose method first and force-fit the question.
Almost always, yes — even anonymous online surveys involving human participants. Health-related surveys may also require HRA approval in the UK.
Cite Guest et al. (2006) on saturation, name your analysis approach, and explain why depth (not generalisability) is the goal. IPA legitimately uses 3 to 8 participants; phenomenology 5 to 15.
Yes with supervisor approval and possibly an ethics amendment. Document the change rationale in the methodology chapter — examiners view methodological adaptation as a sign of researcher maturity, not a weakness.
