What is qualitative analysis?
Qualitative analysis is the systematic interpretation of material such as interviews, open-ended responses, observations and documents. It helps answer questions about experiences, meaning and processes: how people understand a service, what happens during a handoff, or how a program fits into their lives.
The result may be an explanation, a set of developed themes, a case account or a structured comparison. It is more than highlighting quotations or assigning positive and negative sentiment. A useful analysis shows how its interpretation follows from the material and what remains uncertain.
For a growing organization, the practical challenge often starts with feedback arriving from different places. A service team has comments, account history and support notes; a network has local surveys and annual returns. The task is to ask a clear question, preserve the relevant context and review the material in a way that supports a decision. Not every analysis needs a score, a personal identifier or automated coding.
Choose the analytical approach around the question
There is no single procedure that represents all qualitative analysis. Before choosing software or beginning to code, decide what you want the analysis to produce.
Scroll horizontally to see all columns →
| Question | Possible approach | What to watch |
|---|---|---|
| How do people experience joining and using a service? | Thematic analysis to develop patterns of meaning across accounts | Explain which approach to thematic analysis you use and how themes are developed |
| What kinds of issues appear in a defined set of comments? | Content analysis using an appropriate categorization approach | Distinguish category counts from interpretation and population prevalence |
| How do different sites describe the same implementation questions? | A framework matrix that brings cases and analytical categories together | Keep summaries connected to their context; a blank cell is not proof that an issue was absent |
| How does someone describe a change across their life or work? | Narrative analysis focused on the account and its sequence | Fragmenting a story into isolated labels can lose what makes it meaningful |
| How does language frame responsibility or authority? | Discourse analysis | The analysis needs attention to language and setting, not just topic frequency |
These are starting points for selecting a method, not substitutes for learning its requirements. Braun and Clarke distinguish coding-reliability, codebook and reflexive approaches to thematic analysis. Their guidance makes clear that one approach's procedures should not be presented as universal requirements. See Understanding thematic analysis.
This guide uses a small, applied service-review example. It demonstrates organizing evidence and examining an interpretation; it does not claim to be a complete protocol for every research method.
A practical process for an applied qualitative review
1. Write the question and the intended decision
“Analyze our comments” leaves too much undefined. A more useful question is: “How do new customers describe the handoff from onboarding to ongoing support, and what should we investigate before changing that handoff?”
Specify the period, people or settings included and the audience for the result. State whether you are exploring experience, classifying recurring issues, evaluating a process or developing a deeper interpretation. Those purposes affect collection, analysis and the claims you can make.
2. Assemble the relevant material with its context
Keep the question alongside the answer, and retain the date, collection channel and relevant stage of the journey. For interviews, preserve enough of the account to avoid stripping a sentence from its meaning. Check recordings against transcripts where important details are unclear.
Use identifiers only where linking is needed and appropriate. An anonymous service survey can support an aggregate review. Following an account across check-ins requires a suitable account link. A document may need an organization and reporting-period reference rather than a person. The structure follows the question.
3. Read before reducing
Read the material as a whole before relying on labels or summaries. Note early observations, surprises and questions. Separate what someone explicitly said from what you infer. Record your assumptions, including knowledge of the service that could shape your interpretation.
Do not automatically discard short, critical or unusual responses as noise. A brief account may identify something important; an apparently irrelevant answer may reveal that the question was misunderstood. Record exclusions and their reasons.
4. Code or organize the material in a way that fits the method
A code is a label used to organize an aspect of the material. In a structured applied review, a working codebook can record a label, its meaning, inclusion and exclusion guidance, and examples. It can start from the review questions and change as the material reveals something the team did not anticipate.
Keep a history when definitions change. If a revision affects earlier coding, decide whether those earlier records need another review before comparing periods. Other approaches organize analysis differently; do not attach a method label to a process that does not follow it.
5. Examine relationships and develop the finding
Ask what the coded material suggests when read together. Does the same issue mean different things at different stages? Are there accounts that challenge the emerging explanation? Is the apparent pattern a consequence of how the question was asked?
A topic such as “communication” is a useful organizing label, but a finding should say something more specific. For example: “Some customers describe losing a clear point of contact after onboarding.” That statement can be examined against the accounts and other evidence.
6. Check the interpretation and its limits
Return to the full source material. Check whether the selected extracts fairly support the finding and whether contradictory accounts have been considered. Explain how reviewers contributed and how decisions were recorded.
Different methods use different quality practices. Do not apply an agreement statistic, fixed codebook or consensus process automatically to every approach. Braun and Clarke's reviewer guidance cautions against claiming reflexive thematic analysis while describing incompatible procedures.
7. Report what was learned and what should happen next
Describe the question, material, selection, method, finding and limitations. Include enough evidence to make the interpretation understandable without exposing sensitive information. Distinguish a recommended investigation from a proven solution. Assign an owner and a review point when the work is intended to improve an operational process.
Worked example: understanding a service handoff
This is a fictional teaching example. A team reviews twelve voluntary comments about the first month after onboarding. The comments are an exploratory source, not a representative sample of all customers.
Scroll horizontally to see all columns →
| Illustrative comment | Initial observation | Question to investigate |
|---|---|---|
| “The setup session helped, but afterward I wasn't sure who to contact.” | The initial session was useful; the next contact was unclear | Was ongoing ownership explained and easy to find? |
| “I sent the request to two people because I didn't know which team handled it.” | The customer describes uncertainty about routing | Do contact instructions distinguish different kinds of request? |
| “Our contact explained what would happen next, so the change was straightforward.” | A contrasting account describes a clear handoff | What differed in that customer's handoff? |
The team should read all twelve accounts, including comments that do not fit the initial idea. The three extracts alone do not establish a theme or explain every customer's experience. They show how an observation can lead to a more focused investigation.
A cautious finding might be: “Several reviewed accounts describe uncertainty about who owns support after onboarding; a contrasting account describes a clear explanation of the next step. We should examine how ownership is communicated.”
The next step could combine a review of contact instructions with a small test of a clearer handoff message. Check subsequent experience and operational records before deciding whether the change helped. Do not conclude from these comments that unclear ownership caused cancellations.
If you count categories, define what you counted
Suppose four of the twelve comments mention uncertainty about the next contact. That is four of twelve reviewed comments, or about 33%. It is not evidence that 33% of customers experienced the issue. People who respond may differ from those who do not, and one person may have submitted more than one comment.
Report the unit: passages, comments, interviews, people or accounts. A comment may receive more than one category, so category percentages can add to more than 100%. A rare account can also matter greatly; frequency alone does not determine importance.
When combining comments and ratings, preserve their relationship where the design supports it. A comment submitted with a low rating provides that respondent's account of their experience. It does not automatically isolate the cause of the rating. See mixed-methods data analysis for combining evidence without overstating the result.
Make recurring analysis comparable without forcing identical surveys
Different branches, sites and member organizations may need different questions. Agree on the limited shared context needed for the central review: the relevant period, service stage, issue definition and other fields that support the intended comparison. Allow local questions to address local needs.
Document mappings and differences. An open invitation to describe any experience does not produce the same material as a question specifically asking about delays. Translating different wording into one category does not remove that collection difference. Separate results when a combined comparison would mislead.
Maintain stable reference information once where appropriate and update changing context with its date. This helps avoid repeatedly collecting the same details and prevents today's account or site profile from silently replacing the context of an earlier response.
How Sopact reduces coding and reporting work
Human judgment belongs in the definitions and interpretation. Repeatedly applying those definitions across thousands of responses is a different kind of work—and a major source of avoidable labor.
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.
What software should help you do
A small review can be carried out carefully with documents and a spreadsheet. A larger or recurring workflow may benefit from software that reduces repeated preparation, keeps evidence accessible and makes review decisions easier to follow. Tool choice depends on the method, volume, collaboration and governance needs.
Test the actual workflow with a representative sample:
- Can you read an extract in its original context?
- Can you keep relevant dates, questions and record relationships without requiring unnecessary personal details?
- Can you record analytical definitions, revisions and reviewer decisions?
- Can you distinguish a draft interpretation from a reviewed finding?
- Can you compare periods while seeing changes in collection or coding?
- Can the right people access the evidence while others receive an appropriate summary?
- Can you export the material needed to explain the result?
Sopact's fit should be assessed around recurring collection, connected context, analysis and governed use of the result. Ask for a demonstration with your own workflow and verify the controls you need. The relevant distinction is how much ongoing reconciliation and review the team must manage, not a claim that every other tool separates numbers and text.
AI can assist with organizing material and drafting candidate categories or summaries. It can also miss nuance, overgeneralize or produce an unsupported interpretation. A codebook does not guarantee identical or unbiased model output. Keep human responsibility for the analysis, review the underlying material and document how automation was used.
Turn the analysis into a useful report
Use a short structure: the question, what was reviewed, how it was analyzed, what was found, what complicates the finding, and what happens next. Include selected extracts only when sharing them is appropriate. Removing a name may not prevent identification in a small organization.
For a broader reporting workflow, use How to Write an Impact Report and browse report examples. Start with the qualitative-data guide if you are still choosing the material to collect.
Watch: a related qualitative-analysis workflow
This video introduces a connected approach to qualitative evidence. Use it alongside the method and review considerations above; it does not replace a suitable analytical design.
Unified Qualitative Analysis | What Changes Everything
Frequently asked questions
What is the difference between qualitative data and qualitative analysis?
The data is the material, such as interviews, documents or observations. Analysis is the process of interpreting that material to address a question.
Does every qualitative analysis require a codebook?
No. A codebook supports some approaches. Other approaches develop interpretation through different procedures. Choose and describe a method that fits the purpose.
Is coding the same as analysis?
No. Coding can help organize material, but analysis also involves interpreting patterns, context, differences and implications. A list of labels is not necessarily a developed finding.
Can qualitative analysis include numbers?
Yes, when counting fits the question and method. State what was counted and its denominator. Counts from voluntary comments do not automatically estimate prevalence across the population.
Can AI replace the reviewer?
AI can assist with parts of a workflow, but its output needs review. It does not remove responsibility for method selection, interpretation, confidentiality or the claims made in the report.
How long should qualitative analysis take?
It depends on the question, method, material and depth required. Plan review time as part of the work. Faster organization or coding does not necessarily mean the interpretation is complete.
Choose your next step
ASME
Explore how a membership organization uses qualitative and quantitative stakeholder evidence. The service example above is fictional.
Read the customer story →