When should you use qualitative and quantitative methods?
Use qualitative methods when the question needs an understanding of experience, meaning or process. Use quantitative methods when it needs measurement, numerical comparison or estimation. Use both when combining the evidence will answer something important that either strand alone would leave unresolved.
The choice should follow the question. A team does not need an interview to count completed applications, or a population survey to explore an unfamiliar experience in depth. A mixed approach can be valuable, but it adds work and should have a clear purpose.
Nor is the distinction simply “numbers tell what, words tell why.” Quantitative research can investigate causal questions with an appropriate design. Qualitative accounts can illuminate experience and processes without automatically proving cause. Either kind of evidence can be useful on its own.
How the methods differ
Scroll horizontally to see all columns →
| Planning question | Qualitative approach | Quantitative approach |
|---|---|---|
| What are we trying to learn? | How people experience, interpret or navigate a situation | How much, how often, how measures differ or how they are associated |
| What might we collect? | Interviews, observations, documents and open-ended accounts | Structured measurements, counts, ratings and numerical records |
| How might we select sources? | Selection for relevant experience, context and depth | A design suited to description, estimation, comparison or testing |
| What might the analysis produce? | A developed interpretation, account, set of themes or structured qualitative comparison | Summaries, estimates, comparisons or models |
| What must the report explain? | Selection, context, analytical approach and how the interpretation is supported | Measures, population, coverage, calculations, assumptions and uncertainty |
These are broad distinctions, not rigid borders. A survey can include open-ended and closed questions. An interview can collect factual details. A qualitative analysis can include counts when appropriate. Describe what the work actually does rather than assigning a label based only on the collection tool.
When one approach is enough
A clearly defined operational count
A program manager asks how many eligible participants attended at least one session this month. Accurate attendance and eligibility records may answer that question. Adding a satisfaction question would address a different issue; it would not improve the attendance count itself.
An unfamiliar experience that needs exploration
A membership team does not understand how local representatives prepare annual returns. Interviews and observation of the process may reveal information needs and practical difficulties. A large fixed-response survey designed before that exploration could ask the wrong questions.
A specific usability problem
A service team wants to see where people struggle to submit a request. Observing people attempt the task may be more useful than asking only whether they are satisfied. The team can add measurement or interviews if those address a further question, but does not need to add them merely to claim a mixed approach.
A focused, well-executed study is preferable to collecting multiple forms of data that the team has no plan to analyze or use.
When using both adds value
Combining methods is useful when the relationship between the sources will change the interpretation or decision. Write that relationship before designing the collection.
Scroll horizontally to see all columns →
| Decision | Numerical evidence | Qualitative contribution | What combining them could add |
|---|---|---|---|
| Improve a member return process | Completion and missing-field patterns by period | Representatives' accounts of preparing and submitting returns | A more focused investigation of where the process creates work or confusion |
| Review a training program | Participation and suitable follow-up measures | Accounts of trying to use the learning in practice | An understanding of the conditions surrounding observed outcomes |
| Improve a customer handoff | Defined measures of timing and submitted feedback | Accounts of finding the right contact and getting help | A better specified change to test |
| Review partner reporting | Coverage and comparable delivery indicators | Explanations of local conditions and documentation issues | A distinction between a delivery issue, a definition issue and a collection gap |
Each example is a possible study plan, not a claim that collecting both will establish causal impact. The team still needs an appropriate design, reliable sources and a careful interpretation.
Decide how the strands will connect
You might explore experiences first to improve a later instrument, use interviews to investigate a survey result, or collect related sources during the same period. The sequence depends on the question. Fetters, Curry and Creswell describe exploratory sequential, explanatory sequential and convergent designs, along with ways to integrate work at different levels, in their mixed-methods guidance.
For a practical plan, state what will pass from one strand to the other. Will interview findings inform response options? Will a numerical pattern guide follow-up selection? Will the two analyses be examined together to understand an implementation issue?
Do not assume two results become integrated because they appear on the same slide. Explain what reading them together adds, including whether one challenges the first interpretation of the other. For full design examples, continue to mixed-methods research examples.
Worked planning example: a network's annual return
This is fictional. A central team receives 32 of 40 expected organizational returns by the deadline. Of the 32, eight need clarification about what “people reached” means. The team is considering replacing all local questionnaires with one mandatory form.
The numerical evidence describes collection performance: 80% of expected returns arrived by the deadline, and 25% of received returns needed that clarification. It does not reveal why the issue occurred or whether one identical questionnaire is the best response.
The team interviews representatives from varied local settings, including some who submitted without clarification. It explores how they interpreted the field, what local records they used and what work was involved. Those interviews are selected for relevant experience, not to estimate the percentage of all members who think a particular way.
Suppose the exploratory accounts suggest that local activities differ, while a small number of shared definitions remain unclear. That would support investigating a shared core and clearer guidance rather than immediately removing all local questions. The team should test the interpretation against the full material and try the revised approach before claiming that it solves the problem.
Scroll horizontally to see all columns →
| What is known | What is still uncertain | Next step |
|---|---|---|
| 32 of 40 expected returns arrived on time | Why the eight others did not arrive | Investigate nonresponse separately; do not infer its cause from submitted returns |
| Eight received returns need clarification | Whether the problem is wording, source records or another condition | Review the actual field and representative accounts |
| Local activities vary | Which common definitions support a meaningful aggregate | Agree and test a limited shared core with local contributors |
The two strands help narrow the decision. Neither needs to be stretched into a claim it cannot support.
Do the sources have to come from the same people?
No. An individual-level analysis may need known links between a person's records. A network-level interpretation might compare a survey with interviews from a selected subset or with organizational documents. The appropriate connection follows the question and design.
State the level explicitly: person, account, organization, site, period or broader context. Do not invent a personal link when it is unavailable. Do not collect identifying information merely to make every source fit one format.
If sources are linked, preserve what the link means. An interview with a site manager is not the same as an interview with every participant at that site. It can provide contextual evidence without becoming the participant's own explanation.
Design shared definitions while retaining local questions
Federated organizations often cannot and should not impose one questionnaire for every local activity. Identify the few fields needed for an agreed comparison or aggregate. Document the unit, period, eligibility and calculation, then allow local questions to address local needs.
Keep stable reference data once where appropriate and update changing information with its effective date. A local team should not have to re-enter an unchanged organizational profile every time it provides feedback.
A data dictionary helps contributors and reviewers understand the shared fields. It does not make different populations or instruments equivalent. If an aggregate would hide incompatible definitions, present separate results and explain the difference.
Plan quality separately for each strand
For the numerical evidence, inspect coverage, definitions, missing data, measurement and the assumptions behind the analysis. For the qualitative evidence, assess selection, context, depth, analytical approach and support for the interpretation.
Then examine the combined conclusion. Are the sources addressing the same question? Did the team privilege the evidence that matched its expectations? Does disagreement reveal a limitation or an important difference in experience?
A quotation does not validate a statistical result, and a large numerical dataset does not validate an unsupported qualitative interpretation. For analytical guidance, see qualitative analysis, quantitative data analysis and integrating the analyses.
Keep the work proportionate to the decision
Plan time for collection, preparation, analysis and integration. Adding an interview strand means arranging conversations and interpreting the material, not simply collecting transcripts. Adding a survey means designing and testing measures, reaching the intended group and checking coverage.
Start with the smallest scope that can answer the important question well. That may be a bounded exploratory round followed by a better instrument, or a focused follow-up to an existing measure. Avoid asking people for material that will not be used.
Software and AI may reduce parts of the preparation burden, but they do not remove design and review work. Compare implementation effort and ongoing maintenance without relying on claims that automation makes analysis instantaneous or methodologically complete.
How Sopact reduces coding and reporting work
Use both methods when the question needs them. Then design an affordable recurring workflow, so the qualitative strand is not quietly reduced to a small sample because review time ran out.
A workflow with repeated manual work
- Define from an initial sampleRead material and agree on the codebook.
- Apply it across the datasetCode responses and check the result.
- Revise a definitionReturn to affected material and recode it.
- Reconnect the numbersReconcile coded results with ratings and context, then rebuild the view.
The Sopact workflow
- Your team owns the definitionsDecide what each code means and improve it as you learn.
- Apply coding across the eligible dataAutomate application; people review quality and exceptions.
- Reprocess after a definition changesReapply the revised definition across the configured scope instead of recoding each response by hand.
- Ask across coded text and numbersKeep the response, rating and relevant record context connected; inspect the evidence behind the result.
This compares workflow patterns, not a claim that every research tool requires manual coding or separate files. Some already automate parts of this work; compare the complete cycle.
For this codebook-based workflow, the main saving is repeated application and reconnection—not the removal of human judgment. A changed definition can be reapplied across the configured data while reviewers concentrate on quality, exceptions and interpretation. Coded text stays connected to the relevant ratings and context.
Count the recurring work in ownership cost. Include setup, coding, recoding after revisions, source reconciliation, review and reporting, plus your actual platform and processing expenses. A worked scenario of four cycles of 4,000 responses illustrates 272 fewer annual staff hours; it is an assumption-based example, not a customer benchmark. Existing automation, review needs and implementation effort can substantially change the result.
Adjust the workload assumptions and compare total effort →
A reliable assistant should calculate from the selected records and let a reviewer open the supporting evidence. Check the data scope, definition, denominator and access permissions. Reproducible arithmetic does not make every AI interpretation correct.
Watch: Why Qualitative Analysis Stays Small — And How to Scale It
See why revising a codebook creates repeat work, and how connected coding and quantitative analysis change that workload.
Where a connected workflow helps
Sopact should be evaluated around recurring collection, retained context, reviewed analysis and governed use of evidence. A useful demonstration would follow one real question from collection through a report, including changed definitions, missing responses and an interpretation that needs human review.
The aim is a process that an operating team can maintain as data grows. Verify the features and controls required for your setting. Keeping sources accessible can reduce avoidable reconciliation, but storing them together is not a substitute for an integration plan.
Report why both methods were used
Tell the reader what each strand contributed and what combining them changed. Include selection and coverage, the relationship between sources, important disagreements, limitations and the next action. If one strand did not answer its intended question, say so.
For reporting guidance, use How to Write an Impact Report and browse report examples.
Watch: connecting different kinds of evidence
This companion video introduces a qualitative-evidence workflow. Use the question and design choices above to decide whether and how a combined approach fits your work.
Unified Qualitative Analysis | What Changes Everything
Frequently asked questions
Do I always need both qualitative and quantitative methods?
No. Use both when their integration answers an important question better. One well-chosen approach may be sufficient for a narrower purpose.
Do quantitative methods only answer what, while qualitative methods answer why?
That is too simple. Both can contribute to different kinds of explanation depending on the design. Neither automatically establishes causation.
Which method should come first?
The sequence follows the question. Exploration can inform later measurement, numerical findings can guide follow-up, or related sources can be collected during the same period.
Must both sources be collected in one platform?
No. Integration depends on a clear design and interpretation. Suitable tools and source relationships can support it across systems, although recurring reconciliation may add work.
Can local teams keep their own questions?
Yes. Agree on a limited shared core where comparison requires it, retain local questions and document differences that affect interpretation.
Does using both methods make a finding stronger automatically?
No. Each strand must be well designed and analyzed, and the combined interpretation must be supported. More material can also expose unresolved differences.
Choose your next step
ASME
Explore a membership organization’s use of quantitative and qualitative stakeholder evidence. The annual-return example above is fictional.
Read the customer story →