Qualitative vs Quantitative Research
Date: 2026-08-17
Qualitative tells you why and how; quantitative tells you what and how many. Neither answers the other’s question, and most bad research is one being used to do the other’s job.
Qualitative research produces non-numerical evidence — observations, quotes, recordings — about behaviour and reasoning. Quantitative research produces counts and measurements that can be compared and tested.
| Qualitative | Quantitative |
|---|---|
| Why, how, what if | What, how many |
| Few people, deep | Many people, shallow |
| Observed behaviour | Measured behaviour |
| Generates hypotheses | Tests hypotheses |
| 5–8 participants | Hundreds or thousands |
| ”They didn’t see the delivery cost" | "34% abandoned at the delivery step” |
The pairing that works
Neither is complete alone, and the productive loop alternates:
ANALYTICS 34% abandon at delivery
↓ what
USABILITY TEST they expected free
↓ why delivery, and the cost
appears only at step 3
HYPOTHESIS showing it earlier will
↓ reduce abandonment
A/B TEST did it? by how much?
↓ how many
ANALYTICS new baseline
Quantitative finds the problem; qualitative explains it; quantitative confirms the fix. Skipping the middle step is how teams end up testing solutions to problems they’ve guessed at — Hypothesis Design.
What each cannot do
Qualitative cannot tell you how common something is. Five people struggled with the filter. That is not 100% of users, and no arithmetic on five participants produces a rate.
Quantitative cannot tell you why. A 34% abandonment rate is compatible with a dozen different causes, and choosing between them from the number alone is guessing with a chart attached.
INVALID FROM QUALITATIVE
"4 out of 5 users, so 80% of customers"
"this is the most common problem"
"users prefer the blue version"
INVALID FROM QUANTITATIVE
"they abandoned because the form is long"
"the drop is because they didn't trust us"
← both are hypotheses wearing a
finding's clothes
Where each is strongest
QUALITATIVE
early, before anything is built
understanding an unfamiliar audience
diagnosing a known problem
anything involving intent or emotion
— User Interviews, Usability Testing
QUANTITATIVE
sizing a problem
prioritising between problems
measuring whether a change worked
monitoring over time
— Funnel Analysis, A-B Tests
See: User Interviews · Usability Testing · Funnel Analysis · A-B Tests
The counting trap
The most common misuse is counting qualitative data:
“3 of 8 participants mentioned the delivery cost, so it affects 37.5% of users.”
It affects an unknown proportion. Eight non-randomly-selected people are not a sample from which a rate can be estimated — the interval around that 37.5% is so wide as to be useless, and the participants weren’t randomly drawn in the first place — Sample Size in Qualitative Research.
Report qualitative findings as existence, not frequency: “participants missed the delivery cost until checkout” — a real problem worth fixing, with no false precision attached.
Where the line blurs
Some methods produce both:
- Surveys with closed questions are quantitative; open-text answers are qualitative and need coding — Surveys, Qualitative Coding
- Session replay is qualitative observation at quantitative scale, which is genuinely useful and genuinely easy to over-read — Session Replay
- Usability testing with task success rates produces small-sample numbers that should be treated as directional, not as measurements
Combining them properly is Triangulation — when independent methods agree, confidence rises much faster than either method alone justifies.
The practical rule
Ask which question you have before choosing a method. “Is this a problem?” is quantitative. “What is going on?” is qualitative. Teams that own only one of the two capabilities reliably answer the question they can, rather than the one they have.