Fix UX Funnels Faster: 8–10 Qualitative Interviews for Product Teams

Qualitative data is descriptive, non-numeric information that captures qualities, meanings, and lived experiences rather than quantities. It shows up as text, images, audio, video, field notes, and physical artifacts, and it exists to answer “how” and “why” questions that numbers alone can’t touch. Researchers collect it through interviews, observation, and open-ended questions, then interpret it through coding and thematic analysis rather than statistical formulas.

Qualitative Data Meaning: The Forms It Actually Takes

The Australian Bureau of Statistics puts it plainly: qualitative data measures “types” and can be represented by a name, a symbol, or even a number code. That last part trips people up constantly, so it’s worth sitting with for a second.

Qualitative data isn’t defined by what it looks like on a page. It’s defined by what you can legitimately do with it. A researcher might label participants “1 = Male, 2 = Female” in a spreadsheet, and that spreadsheet cell holds a number. But you cannot average those numbers, calculate a standard deviation from them, or run a t-test on them, because the digit is a name tag, not a measurement. This is the difference between nominal data (unordered categories like eye color or job title) and ordinal data (ranked categories like “satisfied, neutral, dissatisfied”), both of which fall under the qualitative umbrella even when they’re written down as numerals.

In practice, qualitative material comes in several recognizable forms that help researchers understand user behavior to optimize AI content strategies:

• Transcripts — verbatim records of interviews or focus groups, often the richest single source in a study

• Field notes — a researcher’s real-time observations, hunches, and contextual details from a site visit or usability session

• Images and video — photographs, screen recordings, or session replays that show behavior instead of describing it

• Audio recordings — raw voice data before or after transcription, useful when tone and hesitation carry meaning

• Diaries and journals — participant-generated logs written over days or weeks, capturing change over time

• Artifacts — physical or digital objects (a sketch, a product mockup, a support ticket) that reveal how people actually use something

The UK Data Service describes this material as giving “voice to the lived experience,” which is really the whole point. A growing share of studies now mix formats deliberately, pairing interview transcripts with screen recordings or diary entries with photos. That multimodal approach demands more from the analyst, since a theme has to hold up whether it’s expressed in a sentence or a facial expression on video.

How to Collect Qualitative Data: Methods That Match Your Question

Picking a collection method before you know your research question is backwards. The question should dictate the method, and the four most common approaches each serve a distinct purpose.

1. Structured interviews work when you need comparability across many respondents. Every participant answers the same fixed questions in the same order, which trades depth for consistency.

2. Semi-structured interviews blend a core question guide with room to follow interesting tangents. This is the workhorse method for most applied research because it balances comparability with discovery.

3. Unstructured interviews let the conversation go wherever the participant takes it. Use this when you’re exploring a topic nobody has mapped yet and don’t want your own assumptions to narrow the conversation.

4. Focus groups surface how people react to each other’s opinions, not just their own. Group dynamics can produce insight you’d never get one-on-one, but a dominant voice can also flatten the discussion, so a skilled moderator matters more here than in any other method.

5. Observation and ethnography capture what people do rather than what they say they do, which are often two different things. Field notes from this method tend to be the most time-intensive to collect but the hardest to fake.

6. Diaries and participant-generated artifacts track experiences as they unfold, avoiding the memory distortion that creeps into a retrospective interview.

7. Open-ended survey items and document analysis scale qualitative inquiry to larger samples, though at the cost of the follow-up questions an interview allows.

Sample size in qualitative research isn’t about hitting a number. It’s about reaching saturation, the point where new interviews stop producing new themes. Most researchers get there through purposeful sampling, deliberately selecting participants who can speak to the research question rather than recruiting at random.

Pro Tip: Run your first two or three interviews as a pilot, then re-read the transcripts before scheduling the rest. If you’re not hearing anything new by interview eight or ten in a focused study, you’ve probably reached saturation and can stop recruiting.

Qualitative Data Analysis: Turning Transcripts Into Findings

Raw transcripts are not findings. They’re inputs. Getting from a pile of interview recordings to a defensible conclusion requires a deliberate process, and skipping steps is where a lot of student research goes wrong.

Coding is the foundation. Analysts typically move through three stages: open coding (labeling raw text with descriptive tags), axial coding (grouping those tags into broader categories), and selective coding (identifying the core theme that ties everything together). From there, a few analytic traditions take over. Thematic analysis looks for patterns of meaning across a data set. Content analysis counts and categorizes the presence of specific words or concepts. Narrative analysis treats each participant’s account as a story with structure, asking how events are sequenced and framed rather than just what was said.

None of this happens in one clean pass. Interpretation is iterative: an analyst reads the same transcript multiple times, and the meaning that emerges the fourth time often looks different from the first. This is where reflexivity comes in. The researcher is not a neutral instrument. Their background, assumptions, and prior exposure to the topic shape what they notice, and good qualitative work names that influence instead of hiding it.

Credibility isn’t optional, and the George Washington University library guide frames analysis as an active interpretive act, not a summary exercise, one that demands methodological transparency to hold up under scrutiny. A few techniques carry that weight:

• Triangulation — checking a finding against multiple data sources or methods to see if it holds

• Peer debriefing — having a colleague challenge your codes and interpretations before you finalize them

• Audit trails — documenting every analytic decision so another researcher could trace your reasoning

• Thick description — writing findings with enough contextual detail that a reader can judge the interpretation for themselves

Qualitative “variables” aren’t variables in the statistical sense. A code like “frustration with checkout” is a conceptual label built through interpretation, not a measured quantity, so running a chi-square test or calculating a mean on coded categories misapplies a mathematical tool to material it was never designed for.

That distinction matters more than it sounds. Treating a code frequency count as if it were a statistically significant result is one of the most common errors in applied qualitative work.

Qualitative vs. Quantitative Data: When to Use Each

Quantitative data answers “how many” and “how much.” Qualitative data answers “why” and “how.” Coursera’s guidance on this split is useful here: qualitative data explains why people act a certain way, while quantitative data tells you how often they do it, and the two are rarely substitutes for each other.

The practical differences run deeper than the question type:

• Data form: numbers and measurements versus words, images, and observations

• Sample logic: quantitative research favors large, randomized samples for statistical power; qualitative research favors smaller, purposefully selected samples for depth

• Analytic technique: statistical tests and regression models versus coding, thematic analysis, and narrative interpretation

• Output: a p-value or confidence interval versus a theme, a typology, or a rich case description

Neither approach beats the other across the board. Quantitative data is stronger when you need to measure scale, compare groups, or track a metric over time. Qualitative data is stronger when you need to understand motivation, uncover a problem you didn’t know existed, or explain a number that doesn’t make sense on its own.

Mixed-methods designs exist because most real questions need both. An explanatory sequential design runs the numbers first and follows up with interviews to explain the pattern. An exploratory sequential design does the reverse, starting with open-ended interviews to build a framework, then testing it quantitatively at scale. A convergent design collects both at the same time and compares them side by side.

Here’s a concrete version of that last pattern: a product team notices click rates on a signup button are strong, but conversion drops right after. The click data tells you where people are dropping off. It doesn’t tell you why. Pairing that funnel metric with a handful of qualitative UX interviews usually surfaces the actual friction, whether it’s a confusing form field or an unexpected pricing surprise, in a way the numbers alone never could.

Advantages and Limitations of Qualitative Data

Qualitative research earns its place in a study for reasons quantitative data can’t replicate. It captures context that a survey scale strips away, and it’s one of the few methods that lets a researcher discover a question they didn’t know to ask going in. Participants get to speak in their own words rather than selecting from a predetermined list, which is exactly why the UK Data Service calls this kind of data a voice for lived experience.

The trade-offs are real, though, and pretending otherwise undermines the credibility of the work.

• Generalizability is limited. Findings from twelve interviews describe those twelve people’s experiences well, not a population.

• It’s time-intensive. Transcription, coding, and iterative interpretation take far longer per participant than a survey response does.

• Interpretation carries subjectivity. Two analysts coding the same transcript can reasonably land on different themes if their frameworks differ.

None of these limitations disqualify the method. They just require honesty. The fix is procedural: state your sampling rationale explicitly, document your coding decisions in an audit trail, and report your findings as context-bound rather than universal. A study that says “these findings reflect the experiences of the sixteen users interviewed” is more credible than one that overreaches, not less.

Concrete Examples Across Disciplines

Abstract definitions only go so far. Here’s how qualitative data moves from raw material to a decision in three different fields.

1. UX research. A researcher watches five users attempt to complete checkout during a usability test, recording field notes on where each person hesitates. Coding those notes reveals a recurring pattern: users pause at the shipping cost reveal, not the payment form itself. The recommendation that comes out of it is specific, showing shipping costs earlier in the flow, not a vague call to “simplify checkout.”

2. Health research. A team collects interview excerpts from patients managing a chronic condition, asking open-ended questions about daily routines. Thematic coding surfaces a shared explanation for medication non-adherence: it’s not forgetfulness, as clinicians assumed, but the timing conflicting with work schedules. That reframes the intervention entirely, from reminder apps to flexible dosing guidance.

3. Market research. Open-ended responses on a customer survey, ones that ask “what almost stopped you from buying,” get coded into recurring concerns. Those clusters become the backbone of a customer persona or journey map, giving a marketing team language pulled directly from customers instead of guesswork.

In each case, the raw qualitative data (a note, a transcript, a sentence) is worthless until someone codes it, and worth very little coded without a decision attached to it.

Tools and Resources for Qualitative Work

Most practitioners lean on a handful of tool categories: transcription software to convert audio into text, multimedia management systems to organize images and video, and qualitative data analysis software (often shortened to QDAS) to support coding and theme-building across large transcript sets.

Product teams have their own version of this stack, and it looks a little different. Instead of interview transcripts, the raw material is session-level behavior, and instead of a QDAS tool, the analysis often happens inside a session replay platform.

• Session replay shows exactly where a user hesitated, misclicked, or abandoned a flow, functioning like field notes for digital products

• Heatmaps and click maps aggregate that behavior visually, turning individual sessions into a pattern

• Error tracking connects a qualitative observation (a user got stuck) to a technical cause (a broken form validation)

LiveSession’s positioning example: Livesession pairs these session-level qualitative observations with quantitative engagement metrics in one dashboard, so a product manager can watch the heatmap and click behavior that explains a drop in a funnel metric without switching tools. The platform also builds in GDPR and CCPA compliance and connects with tools like Intercom, Zendesk, Shopify, and Segment, which matters when session recordings potentially contain sensitive user data.

Reporting Qualitative Findings Responsibly

How you report qualitative data matters almost as much as how you collect it. A quote pulled out of context can misrepresent a participant’s actual position, so responsible reporting pairs quotes with enough surrounding detail (thick description) that a reader can judge the interpretation rather than take it on faith. Findings should also state their limitations plainly rather than implying broader applicability than the sample supports.

Ethics starts before data collection even begins. Recorded interviews need informed consent that specifically covers recording, storage, and eventual use. Anonymization should happen early, stripping names and identifying details before a transcript circulates beyond the core research team. Secure storage isn’t a formality when the data includes health information, financial details, or anything covered under GDPR or CCPA.

• Get explicit consent for recording before the interview starts, not after

• Anonymize transcripts at the point of transcription, not as an afterthought

• Store raw recordings separately from de-identified analysis files

• Decide in advance what gets shared as excerpts versus what stays internal

Pro Tip: If you’re sharing findings publicly, share coded themes and redacted excerpts, not full transcripts. Full transcripts almost always contain identifying details that anonymization misses on a first pass.

Why Qualitative Data Changes Product Decisions

They never tell you why. That gap is where onboarding gets fixed, where a confusing label finally gets rewritten, where a “successful” feature that nobody actually likes gets caught before it ships broadly. Teams that treat qualitative and quantitative data as one integrated view, rather than two separate reports, make faster and better product calls. If you want to see what that combined view looks like in practice, Livesession’s guide to qualitative UX research and its categorical versus quantitative data breakdown are worth the read.

Sources

For deeper grounding, the Australian Bureau of Statistics explains the categorical-variable framing behind qualitative labels. The UK Data Service covers common data forms in depth. George Washington University’s library guide walks through analysis and rigor, and CASRAI offers a clear reference on coding terminology.

• Quantitative and qualitative data | Australian Bureau of Statistics

• Overview - Qualitative Data: Best Practices, Analysis, and Tools - Research Guides at George Washington University

• Qualitative data - UK Data Service

FAQ

What Is the Simplest Definition of Qualitative Data?

Qualitative data is non-numeric, descriptive information (words, images, audio, video, or observations) that captures meaning and experience rather than measurable quantity. It answers “how” and “why” questions and typically requires coding or thematic analysis instead of statistical tests.

What Is Meant by Qualitative Data in Statistics?

In statistics, qualitative data refers to categorical variables, ones that describe types or classes rather than measurable amounts. The Australian Bureau of Statistics notes these can appear as names, symbols, or number codes, but the labels aren’t meant for arithmetic.

How Is Qualitative Data Different From Quantitative Data?

Qualitative data is descriptive and non-numeric, gathered through interviews or observation and analyzed by coding and theme-building. Quantitative data is numeric, gathered through measurement or counting, and analyzed with statistical methods. One explains why something happens; the other measures how much or how often.

What Are the Main Methods for Collecting Qualitative Data?

The most common methods are interviews (structured, semi-structured, or unstructured), focus groups, observation or ethnography, open-ended survey questions, and participant diaries. The right method depends on whether you need depth from a single respondent or breadth across many.

Can Qualitative Data Include Numbers?

Yes, but only as labels, not measurements. A code like “1 = Male, 2 = Female” is qualitative because the numeral represents a category, and running arithmetic on it (an average, for instance) produces a meaningless result.

How Do Researchers Analyze Qualitative Data?

Researchers typically code the raw material (open, axial, then selective coding) and group codes into themes using thematic, content, or narrative analysis. Rigor comes from techniques like triangulation, peer debriefing, and audit trails, as outlined in George Washington University’s research guide.