Blog Marketing
Qualitative vs. quantitative research: marketing examples

The difference between qualitative and quantitative research comes down to the question and the conclusion you need. If you do not yet know why people leave checkout, listen to their experiences and examine the journey. If you need to estimate where it happens most often, measure events using consistent definitions. If you need both answers, design a sequence that connects the methods.
What each method can tell you
The University of Texas at Arlington Library describes qualitative research as including in-depth interviews, focus groups, and document reviews to identify themes; it can help develop explanations or hypotheses from the cases studied. Quantitative research includes surveys, structured interviews, measurements, and numeric records; it can help analyze patterns and examine stated hypotheses. The guide also contrasts depth across a few cases with breadth across more cases (UTA’s comparison of qualitative and quantitative methods).
That is an introductory map, not a rigid boundary. A structured interview can produce answers that are coded numerically; a survey can include open-ended questions. The design and analysis determine what kind of evidence a study produces.
On smaller screens, swipe the table horizontally to read every column.
| Marketing need | Approach that may help | Typical evidence | Conclusion it can support |
|---|---|---|---|
| Understand what confuses or concerns buyers | Qualitative | Interviews, journey observations, support comments | Possible explanations and the language of people interviewed |
| Measure where people leave checkout | Quantitative | Analytics events with defined stages | Frequency or proportion observed in the available data |
| Estimate how many people share a view | Quantitative, with an appropriate sample and instrument | Survey of a defined population | An estimate whose generalizability depends on sampling and response |
| Find out whether a change caused an improvement | An experiment designed to compare outcomes | Comparable measurement and assignment, with relevant factors controlled | An effect attributable under the experiment’s conditions |
The categories do not guarantee a sound answer. A large event log can measure the wrong thing if people are missing, sessions are duplicated, or “abandonment” has no stable definition. A handful of interviews can reveal a specific obstacle without showing how widespread it is.
A worked example: checkout abandonment
Imagine an online store sees many sessions with items in the cart and few completed purchases. This scenario is hypothetical: it illustrates research decisions and does not describe a real store’s results.
Question. Separate two questions: “At which checkout step do more sessions stop?” and “What prevents some people from continuing?” The first asks for a pattern; the second asks about experiences and reasons. Neither is answered by counting abandoned carts until abandonment and the cases included are defined.
Selecting cases. To explore reasons, invite people who recently started checkout but did not buy, aiming to include different journeys such as mobile and desktop, and different steps. Do not select them only because they responded to an ad. Seek variation that helps compare experiences, not a miniature statistical version of the customer base. Use neutral prompts—“Tell me what happened after you added the item”—and let people describe the journey before suggesting possible causes.
To measure the pattern, define the population and time window (for example, checkout sessions started during four weeks), specify the start and completed-purchase events, and check that they are recorded consistently. If you use a survey, define whom the invitation frame represents, track nonresponse, and do not treat volunteers as though they were randomly selected.
Analysis. For the qualitative part, organize responses into themes found in the material—for example, surprise at shipping cost, uncertainty about delivery date, or an address-validation error—and retain cases that do not fit. These are candidate explanations, not population frequencies. For the quantitative part, calculate how many valid checkouts reach each step and where the proportion drops; compare device or region only when definitions and volume support an interpretable comparison. Check for duplicates, missing events, and instrumentation changes.
Decisions. Suppose, only for illustration, interviews suggest shipping cost appears too late, and analytics show a drop at the step where it is revealed. The store could test an earlier estimate, explain delivery terms, and measure progression to the next step and completed purchases. If the issue appears only on one device, review that flow. If accounts point to several causes, test separate changes instead of attributing the drop to one explanation. To claim a change caused an increase, an appropriate experimental comparison would be needed; improvement afterward alone could also coincide with seasonality, campaigns, or other changes.
The sequence gives each method a different job: explore to understand possibilities, measure to size a pattern, and experiment to evaluate causality. If quantitative data conflict with interviews, do not immediately discard either; check whether interviewees represent a particular situation, whether the metric captures the friction, and whether both measurements cover the same period.
A large sample does not fix biased selection
Size and representativeness answer different questions. A survey with thousands of voluntary responses may overrepresent people with more interest, time, or an unusually strong experience. A smaller sample selected from a clear frame may suit a bounded question; even then, precision and generalizability depend on design, coverage, and nonresponse. Qualitative research often aims to develop detailed understanding of experiences, not estimate percentages for an entire population.
Sample size also does not establish causality. Observing that people who see a promotion buy more does not prove the promotion caused it: they may have arrived through another channel or already had stronger purchase intent. Causal evaluation needs a design that validly compares conditions and addresses alternative explanations, such as a well-implemented experiment when feasible.
How to decide what to do first
- Write down the decision. “Which message should we change?” calls for understanding language and friction; “what share abandons on mobile?” calls for a measure with defined events and a population.
- Separate what you know from what you infer. A recorded click is an event; “shipping cost scares people off” is an explanation that needs evidence.
- Choose cases and instruments to match the conclusion. To understand experiences, find informative cases and listen without leading. To estimate reach, define the population, selection, measurement, and treatment of nonresponse.
- Combine methods when one answer enables the next. Interviews can help shape survey options; measurement can show where to investigate further; an experiment can evaluate a specific intervention.
- State the limit when reporting results. Explain what was observed, who was included, and which decision it supports. Do not turn a percentage from a convenience sample into a claim about all customers.
The focus group guide explains one qualitative group technique; the customer journey map helps represent stages in an experience. A technique can contribute evidence to a study, while a map organizes the experience you want to investigate.
Plan your next post in HeyMark.
Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.