What is qualitative data?
Qualitative data records qualities, experiences, practices and meaning. It can include interview transcripts, open-ended survey answers, field notes, documents, photographs and recordings. Researchers and operating teams use it to understand how people experience a situation, how a process works and what a numerical measure may overlook.
It is useful in its own right. It does not have to become a count or attach to a named person before it can support a finding. What matters is whether the material, collection approach and analysis fit the question, and whether the interpretation is supported by the evidence.
For example, an association might explore how member organizations use its resources, a service team might investigate confusing handoffs, or a training provider might examine what helps learners apply a skill. Each question calls for context and a careful reading of experience.
Types of qualitative data, with examples
The following are useful ways to organize material, rather than mutually exclusive categories. A recorded interview, for instance, may also have a transcript and field notes.
Scroll horizontally to see all columns →
| Material | Example | Context to preserve |
|---|---|---|
| Text | An open-ended survey response, transcript or case note | The question, source, date and relevant surrounding account |
| Observation | Notes about how people navigate a service process | The setting, observer, observation approach and distinction between description and interpretation |
| Audio and visual material | A recording, photograph or video of an activity | Capture conditions, permissions and relevant time or location information |
| Documents | A policy, meeting record, report or correspondence | Author or source, purpose, audience, version and period covered |
A transcript may lose pauses, tone or visual context that matters to some questions. A photograph may show a condition without explaining why it occurred. Choose the material and preparation process around what the analysis needs to understand.
How is it different from quantitative data?
Quantitative data describes quantities or measurements numerically. Qualitative evidence explores meaning, experience and processes through material such as words and images. The distinction is not “numbers tell what, comments prove why.” Both kinds of evidence require a suitable design and interpretation.
A count of missed appointments may reveal a pattern. Interviews may help investigate how scheduling, communication or other circumstances were experienced. Neither source alone automatically establishes the cause of every missed appointment.
Combining the sources can be useful when the question calls for it. They do not always need to come from the same people: interviews with a purposively selected group can add context to a wider survey, as long as the relationship between those samples is explained. See quantitative data and mixed-methods analysis for the related distinctions.
How to collect useful qualitative data
Start with the learning question
Decide what you need to understand. “What happens when a member organization tries to submit its annual return?” invites a different conversation from “Are members satisfied?” Choose the people, settings and sources that can illuminate the question, including experiences that may challenge your assumptions.
Choose the collection approach
Interviews allow follow-up questions; focus groups can reveal interaction and shared or contested views; open-ended surveys provide a written response without an interviewer present. Observation and document review answer other questions about practice and context. A case study can combine several sources around a bounded case.
These approaches have different demands and limitations. Sensitive personal experiences may be difficult to discuss in a group. A short survey comment may identify a problem without explaining its history. See qualitative data collection methods for a fuller comparison.
Ask and record carefully
Use clear, open questions and neutral follow-ups. Ask for a concrete experience where useful. “Tell me about the last time you tried to complete the return” may produce more useful detail than asking someone to endorse the team's preferred explanation.
Preserve enough context to interpret the response, while limiting unnecessary personal information. Explain how the material will be used and who can access it. Check recordings and transcripts rather than assuming transcription is error-free.
How to analyze qualitative data
There is no single required procedure for all qualitative work. Choose the approach that fits the research question, the material and the kind of finding you intend to develop.
Scroll horizontally to see all columns →
| Approach | What it can focus on | Important choice |
|---|---|---|
| Thematic analysis | Patterns of meaning across a dataset | Specify the particular approach and how themes will be developed |
| Qualitative content analysis | Systematic interpretation and categorization of content | Explain the categories, unit of analysis and role of any counts |
| Framework analysis | Comparisons across cases and topics in an organized matrix | Keep summaries connected to the underlying accounts |
| Narrative analysis | How accounts are told and structured over time | Preserve the story and context rather than reducing it immediately to fragments |
| Discourse-oriented analysis | How language constructs meaning and social relationships | Use a method and level of detail appropriate to the question |
Even within thematic analysis, procedures differ. Braun and Clarke distinguish coding-reliability, codebook and reflexive approaches; reflexive thematic analysis involves active interpretation and theme development rather than requiring a fixed coding scheme applied by interchangeable coders. Their guide to understanding thematic analysis explains why those choices matter.
For an applied feedback review, a practical sequence might be to read the material, note initial observations, develop or refine categories, examine variation, check interpretations against the sources and write the finding. Document how you worked. Do not label a procedure as one method while applying incompatible rules from another.
A worked example: understanding a difficult member return
Consider a fictional association reviewing annual reporting. It interviews 12 representatives selected to include different organization sizes and reporting experiences. The team wants to understand the work involved in preparing the return, not estimate the percentage of all members who have a problem.
Three illustrative excerpts might read:
“The form was clear. Finding the figures from three local teams took the time.”
“We counted visits, but the question asked for people. I wasn't sure what to submit.”
“We sent everything in and never heard what the network learned.”
These fictional excerpts suggest distinct issues: gathering local evidence, interpreting a measure and receiving something useful after contributing. Calling them all “negative sentiment” would lose much of their practical meaning.
Scroll horizontally to see all columns →
| Initial observation | What to examine in the wider material | Possible operational response |
|---|---|---|
| Preparation involves several contributors | How information moves between teams and where delays occur | Clarify contributor roles and the preparation schedule |
| Measure definitions differ | Whether people and visits are consistently distinguished | Agree a shared definition or report different units separately |
| Contributors do not see a return from reporting | What feedback they expected and what was actually provided | Plan an appropriate member summary and discussion |
These are starting interpretations, not final themes established by three quotations. Read the full interviews, look for conflicting accounts and consider whether the examples fairly represent the material. Record the reasoning behind the final finding.
A defensible report might describe the preparation burden among the interviewed representatives and explain how they were selected. It should not claim that the same proportion of the entire membership experiences that burden.
Can you count qualitative data?
You can count coded observations when that serves the question and the method supports it. Define what is being counted: people, responses, documents or passages. One person mentioning a topic five times is not five people experiencing it.
For example, a fictional review of 80 usable comments might identify scheduling concerns in 24 responses. That is 30% of the reviewed comments, assuming each response is counted once for that category. It is not automatically 30% of all participants or members. State whether responses can have more than one category.
Frequency is not the only measure of importance. A rare account may reveal an exclusion, serious problem or overlooked experience that deserves attention. Some qualitative approaches develop findings without converting them into prevalence estimates at all.
What makes qualitative evidence trustworthy?
Show how the question, sampling, collection and analysis fit together. Preserve source context, explain analytical decisions and consider the researcher's or reviewer's role in interpretation. Include contradictory evidence where it changes the finding.
A codebook can help an applied team communicate category meanings. It may change as the team learns; record changes and decide whether earlier material needs another review. A permanently fixed codebook is not a universal requirement for qualitative quality.
Use quality checks suited to the chosen approach. Agreement between coders may matter for some coding tasks, but it is not the universal standard for all qualitative research. A repeated theme distribution is also not proof that the interpretation is adequate.
Keep context without requiring everyone to be identified
Some workflows need an authorized continuing record, such as case follow-up. Others need anonymous feedback, organizational context or a document-level source reference. A policy document does not have to be attached to a participant to be analyzable.
Across a network, agree on the small set of shared fields needed for useful comparison. Local teams can keep questions relevant to their own work. Record changes in questions and definitions so differences between periods are not mistaken for changes in experience.
Keep access appropriate to the material. Names removed from a quotation may not make it anonymous if the situation is distinctive. Review quotations and exports for what they could reveal to their intended audience.
How Sopact reduces coding and reporting work
A larger response set should not force the team to freeze an early codebook or ignore material that does not fit it. The cost of applying revised definitions matters.
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 software and AI can help
Software can organize sources, support coding and retrieval, connect relevant context and maintain a record of analysis. Manual analysis is not limited to a few dozen responses by a universal rule; the effort depends on the material, method, team and required depth.
AI can assist with tasks such as suggesting categories or locating passages. Check omissions, ambiguous wording, multilingual meaning and contradictory evidence. A fixed instruction or codebook does not guarantee identical AI outputs, and consistent output does not guarantee a sound finding.
Sopact's approach combines recurring collection, connected source context, reviewed analysis and governance. Test how it supports your actual workflow and analytical method. Keep the distinction between an AI suggestion, a reviewed interpretation and an approved report clear.
Watch: unified qualitative analysis
This related explainer discusses bringing qualitative evidence into a connected review workflow. Apply the method and interpretation checks above to your own material.
Present findings with enough context to understand them
Explain the question, sources, selection approach, analytical method and limits. Use quotations to support interpretation, not as decoration or proof of population-wide prevalence. State what the team learned and what action or further investigation follows.
For reporting, use How to Write an Impact Report and browse report examples.
Frequently asked questions
Does qualitative data need a fixed codebook?
No. The requirements depend on the analytical approach. Codebooks can support some applied reviews, while other approaches develop codes and themes through an iterative interpretive process.
Must qualitative data identify the participant?
No. Anonymous responses, documents and group-level context can support analysis. Use identifying links only where they are needed, appropriate and authorized.
Are themes the same as topics or sentiment?
Not necessarily. A topic says what the material concerns; sentiment summarizes an evaluative tone. A developed theme can explain a pattern of meaning. State what your analysis produces rather than treating those labels as interchangeable.
Can qualitative findings be generalized?
The appropriate claim depends on the design and analytical purpose. A purposive interview sample usually does not support a statistical prevalence estimate for the whole population. Explain the setting, selection and relevance of the finding.
Does AI make qualitative analysis unbiased?
No. Collection, models, instructions and reviewer decisions can all affect interpretation. Review the evidence and analytical process rather than treating automation as a guarantee of neutrality.
