Mixed Methods Design: A Worked Example
Sequential explanatory design in practice — survey first, interviews second, with the qualitative sections deliberately mirroring the quantitative ones so the two strands actually meet.
What the design is
The examined mixed methods thesis names its design as sequential explanatory: an anonymous internet questionnaire first, then one-on-one semi-structured telephone interviews to interpret and validate what the survey produced. The word doing the work is explanatory — the second strand exists to explain the first, not to run alongside it.
- Literature → hypotheses
- Survey
- Statistical analysis
- Interviews
- Interpretation
- Revised model
The two strands as they were run
THE TWO STRANDS IN THE EXAMINED THESIS
| Quantitative strand | Qualitative strand | |
|---|---|---|
| Instrument | Anonymous internet questionnaire, nine sections, 7-point Likert | Semi-structured telephone interview, roughly one hour |
| Recruitment | 525 invitations to senior R&D decision-makers across three world regions | 28 experts approached via a professional networking site |
| Achieved | 92 responses; 21 incomplete and rejected; 71 valid | 10 interviews, against a stated target of at least ten |
| Participant profile | 26.2% C-level, 54.8% director-level, 19.0% senior manager | All 20+ years' experience; 80% held doctorates |
| Analysis | Descriptive statistics; one-way ANOVA with post-hoc tests at p = 0.05 across seven relationships; multiple regression with normal P-P plots and residual scatterplots | Manual coding; each interviewee coded with two letters and profiled in a table |
| Sample defended by | Two published participants-per-predictor rules of thumb | A stated target, plus depth of participant experience |
Every figure here is a finding of that one study. None is a standard, a target or a norm.
The device that makes the strands meet
Most mixed methods projects fail at the join. Two strands run, two sets of results are reported, and the discussion asserts that they agree. The examined thesis avoids this with one structural decision.
How results were organised
- Chapter 5 — quantitative results and data analysis
- Chapter 6 — qualitative analysis, sections mirroring chapter 5
- Chapter 7 — discussion, bringing both strands back onto the conceptual model
- Split by method, not by theme — unusual, and it works because of the mirroring
What a themed split would have cost
- The reader could not see which strand produced which claim
- The sequential logic — quant first, qual explains — would be invisible
- Weak survey findings could be quietly propped up by interview quotes
- The separate discussion chapter would have nothing left to do
Construct robustness: one variable, five proxies
The thesis operationalised firm size through five different proxies and tested against each. That is a demonstration of construct robustness — if a relationship holds regardless of how you measure the construct, the relationship is unlikely to be an artefact of the measure.
The model as a moving object
The conceptual model is proposed in the theory chapter, restated in the discussion, then revised twice as evidence accumulates. The reader can watch it change.
- Theory chapterModel proposed, with seven hypotheses derived from the literature that produced it
- Quantitative chapterHypotheses tested; one relationship fails
- Qualitative chapterInterviews interpret the specific findings, section by section against the survey
- DiscussionModel restated, then revised twice against the accumulated evidence
- ConclusionAll seven hypotheses restated verbatim with a verdict appended — not supported, partially supported, supported
The restate-and-adjudicate device in the final row is the most copyable structural feature in the whole exemplar set, and is set out in Where research questions and hypotheses live.
What the thesis got wrong
Applying the pattern
If you are running a sequential explanatory design
- State the order and say why it is that way round — explanatory means qualitative follows and explains
- Give the qualitative chapter the same section structure as the quantitative one
- Split results by method and add a separate discussion chapter to join them
- Operationalise your central construct more than one way and test against each
- Restate hypotheses in the conclusion with an explicit verdict, including the ones that failed
- Put both instruments in the appendix — questionnaire and interview guide
- Keep the methodology chapter about your study, not about research methods in general
The instrument-level detail behind each strand is in Survey design and response rates in practice and Interview design in project research.
What to carry forward
- Sequential explanatory: quantitative first, qualitative second to explain it. The order is the design.
- Mirror the qualitative chapter's sections onto the quantitative chapter's. That single decision is what makes the strands meet.
- Split results by method and let a separate discussion chapter do the joining.
- Measure your central construct several ways. It costs nothing and pre-empts the obvious objection.
- Two-thirds textbook exposition in a methodology chapter is a fault, not a convention — even when examined work does it.
Frequently asked questions
What does sequential explanatory mean?
The quantitative strand runs first and the qualitative strand follows to explain and validate what it found. The alternative orders — qualitative first to build an instrument, or both concurrently — answer different questions. The supplied teaching source does not name any of these designs.
How many interviews are enough?
No supplied source gives a number. The examined thesis set a target of at least ten and conducted ten, defending the sample by the depth of participant experience rather than by size. That was that study's judgement, not a standard.
Should results be split by method or by theme?
The examined thesis split by method and joined the strands in a separate discussion chapter. That works because the qualitative sections mirror the quantitative ones. Splitting by theme without that mirroring tends to hide which strand produced which claim.
Why test one construct with several measures?
If a relationship holds no matter how you operationalise the construct, it is unlikely to be an artefact of the measure. The examined thesis used five proxies for firm size and tested against each.
Do I need coding software for qualitative analysis?
The examined thesis stated that with ten interviews it was not needed and coded manually. What matters is that the decision was stated and justified rather than left for a reader to infer.
References and source attribution
- Five examined research works supplied as exemplars: two doctoral theses (1999, 2016), a doctoral portfolio thesis (2004), a peer-reviewed journal paper (2014) and a conference paper (2017/2018). Structural observations only; chapter bodies were not reproduced.
- Supplied teaching source, Weeks 2-4: Introduction to Research Methods, Developing a Research Topic, Reviewing the Literature.
- Bryman, A. 2016, Social Research Methods, 5th ed., Oxford University Press, Oxford.
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London.
Suggested questions for Ask KEVOS
- Design a sequential explanatory study for my research question.
- How do I structure a qualitative chapter so it mirrors my quantitative one?
- What should go in a methodology chapter that is about my study rather than about research methods?
- Suggest several ways to operationalise my central construct.
- How do I report a hypothesis that was not supported?