Qualitative vs quantitative research: choose what your decision needs
Use qualitative research to investigate experiences and explanations. Use quantitative research to measure defined variables and comparisons. Their order follows the missing evidence: an episode can define what to measure, while a measured pattern can identify whom to interview.
The meal-service team needs to understand what changed between first order and renewal. Interviews can reveal the sequence; a defined measure can show its distribution; a change test can evaluate an intervention. These are different outputs in one fictional teaching case, not competing definitions of rigor.
Choose the output before the research label
All material in this example is fictional. A meal subscription company needs to understand why some customers stop after their first order.
Fictional teaching data
Nonrenewal among two different response bases
Comparison | % |
|---|---|
Schedule changed · 7/12 | 58.3 |
No change · 2/8 | 25.0 |
Twenty fictional records. Schedule-changed group: seven nonrenewers and five renewers; unchanged: two and six. Descriptive rates, not a causal effect.
Design element | Qualitative approach | Quantitative approach |
|---|---|---|
Question | What happened between first delivery and the decision to stop? | How do specified experiences differ between renewers and nonrenewers? |
Audience | Deliberately varied recent customers, including people who stopped | Defined cohort with recorded renewal status and a stated invitation design |
Instrument | Episode interview and packaging walkthrough | Comparable experience questions linked, with permission, to renewal records |
Output | Explanatory account of friction and competing interpretations | Counts, response bases and comparisons on the measured items |
Weakness | Cannot establish how common each mechanism is | May miss explanations absent from the response options |
A report with words can contain counts and a report with numbers can contain explanations. The difference lies in the design: how experiences are selected, what is recorded and what inference the evidence can support. Counting interview themes does not supply a recruitment frame or a population measure.
An episode changes what marketing should ask
Fictional interview excerpt: “The recipes looked manageable, but by Wednesday two ingredients had gone bad. I stopped because I couldn’t predict when I would cook.”
Possible codes: uncertain schedule; ingredient storage; perceived waste. Interpretation: the difficulty may concern flexibility rather than recipe complexity. Counterexample to seek: a customer with an unpredictable schedule who renewed. That person may reveal a workaround or a boundary to the explanation.
A useful output contains the excerpt, context, code and competing account. “Customers want convenience” removes the detail that makes the finding actionable.
The account identifies unpredictable cooking and ingredient waste. Recipe simplicity was not the central problem in that episode. That moves the next question toward commitment timing, storage or quantity.
The continuer with an unpredictable schedule matters because it can reveal a workaround. Without that contrast, the team may build flexibility around a plausible story that successful customers have already solved differently.
Read the twenty-record comparison with its real bases
In a separate fictional demonstration, 20 customers answer “Did your cooking schedule change after placing the first order?” Twelve answer yes. Seven of those twelve do not renew, compared with two of the eight who answer no.
Report 7/12 and 2/8, alongside the full bases and data origin. The demonstration suggests a relationship worth investigating. It does not prove changed schedules caused nonrenewal, and its invented counts provide no evidence about an actual customer population.
If the data were real, you would still check response coverage, missing values, recruitment and other differences between the groups. A small p-value would not repair an unmeasured confounder or missing audience.
There are twelve schedule-changed records and eight unchanged records, so comparing counts alone would obscure the different group sizes. Seven of twelve is about 58%; two of eight is 25%. The roughly 33-point difference is descriptive within the invented data.
It directs a follow-up, not a causal verdict. What other difficulties occurred with schedule change? Which changed-schedule customers still renewed, and how? The five renewers in that group are as informative for the intervention design as the seven nonrenewers.
If the business asks whether a later cutoff changes renewal, implement and evaluate that cutoff under an appropriate comparison. The interview explanation and the cross-tab justify investigating it; they do not supply its effect.
Translate the sequence into separate measures
The fictional ingredient-waste account suggests several variables, not one universal “convenience” score. A schedule changed; ingredients were stored; waste occurred; the customer interpreted the loss as a reason to stop. Specify the period and event for each item before measuring it.
Episode detail | Candidate item | Boundary |
|---|---|---|
Cooking became unpredictable | Did planned cooking days change after ordering? | Does not measure how severe the disruption was |
Ingredients went bad | Were any supplied ingredients discarded before use? | Does not establish why they became unusable |
Customer stopped | Recorded renewal status in a defined window | Does not reveal the cause on its own |
Flexibility might help | Reaction to a complete flexible-scheduling offer | Does not establish its observed renewal effect |
Retain the episode and contrary cases while drafting the instrument. A person whose schedule changed but who renewed may have frozen ingredients, shared meals or used another workaround; the proposed explanation needs room for those differences.
A single convenience rating would merge changed plans, storage, wasted food and the renewal decision. Keep those events separate so the quantitative result can tell the team which action to investigate. Reuse the language learned in episodes, then pretest whether people understand the event and period.
Choose discovery first, measurement first or both
Use discovery first when the response list is unknown. Use measurement first when a defined outcome pattern exists and you need contrasting cases to explain it. Use both together when a decision needs a numerical comparison and a mechanism, aligning audience, period and product version.
For the meal program, the combined brief asks which feasible change manages uncertainty: cutoff, bundle or storage support. The output should show the account, measure, contrary case and action. The receiving team can then commission the implementation test that the chosen commitment needs.
Continue with qualitative guide, quantitative guide, mixed methods example, Discuss a scoped study with Sapien.
