Quantitative research measures what people do at scale: conversion rates, drop-off, clicks and revenue across thousands of sessions.
Qualitative research explains why they do it, through surveys, interviews, session recordings and usability tests with a smaller number of people.
In CRO you need both: numbers show where a page leaks, words and observation suggest what to change, and an A/B test decides whether the change worked.
That split matches the way Nielsen Norman Group defines the two: qualitative studies give "a direct assessment of the usability of a system," while quantitative metrics such as task completion rates give "an indirect assessment."
The rest of this page shows how each side works in conversion research, with real tests from our database where one fed the other.
| Quantitative research | Qualitative research | |
|---|---|---|
| Answers | What, where, how many, how much | Why, what's confusing, what's missing |
| Typical CRO methods | Analytics funnels, A/B tests, click and scroll heatmaps, form analytics | Post-purchase and on-site surveys, customer interviews, usability tests, session recordings |
| Sample size | Hundreds to tens of thousands of visitors | Often 5-8 participants per usability round; dozens to hundreds of open-text survey answers |
| Output | Rates, averages, confidence intervals | Quotes, objections, observed struggles, themes |
| Main weakness | Tells you a step is failing, not why | Can't tell you how common a problem is or whether a fix pays off |
| Wrong choice when | Traffic is too low to measure, or you don't yet know what to test | You need to prove a change raised revenue |
Quantitative UX research: finding where the page fails
Quantitative UX research turns behavior into numbers you can compare.
In conversion work that usually means three sources: your analytics funnel, behavioral tools like heatmaps, and controlled experiments.
Analytics and funnels
A conversion funnel in your analytics tool shows which step loses the most people.
It's the cheapest way to decide where research effort should go, because a page that 2% of visitors reach rarely deserves a week of interviews.
Oliver Kenyon's test for a wild bird food brand started from exactly this kind of number: 70% of visitors were leaving the collection page without opening a single product, and most were first-timers.
The number located the problem. It didn't say what to fix, and the obvious guess (make the products look better) turned out to be wrong.
Heatmaps and scroll data
Heatmaps sit on the line between the two types.
They aggregate thousands of sessions, so they're quantitative, but you read them for patterns rather than proof.
Dave Diederen used heatmaps and scroll behavior to see which product-page sections buyers kept returning to before purchase, then moved those points above the fold.
The A/B test gave +6.23% conversion rate and +7.28% revenue per visitor, with no change in AOV.
In another test, Dave Diederen placed a countdown timer where heatmaps showed collection-page shoppers were already looking: +20.67% conversion rate and +20.65% revenue per visitor.
A/B tests are the quantitative step that settles it
An A/B test is the only research method on this page that shows cause and effect on revenue.
It's also the place where quantitative research is most often misread, because a lift on one metric can hide a loss on another.
Clean Commit's product image carousel redesign is a good example.
Across 84,028 visitors, add-to-cart rose 4.5%, but conversion rate moved only +1.1% and revenue per visitor fell 1.1%.
The usability fix made people click more; it didn't make them buy more.
When quantitative research is the wrong tool
- Low traffic. If a page gets a few hundred visits a month, a funnel report or A/B test can't separate a real change from noise. Talk to customers instead.
- You don't know what to test. Numbers can't generate hypotheses. A page with a 70% exit rate has dozens of possible causes.
- Small-sample usability metrics. Jakob Nielsen's guidance on quantitative usability studies recommends about 20 users, which still leaves a margin of error of roughly ±19% of the mean. Treat task-time figures from a handful of sessions as rough.
Qualitative research in UX: finding out why
Qualitative research collects words and observations. It's how you learn the objections, doubts and confusions that numbers only hint at.
The main methods in CRO are surveys, customer interviews, usability testing and session recordings.
Surveys: the qualitative source behind many winning tests
In our database, post-purchase surveys show up again and again as the starting point for winning tests.
Asking recent buyers what nearly stopped them gives you objections in the customer's own words, which you can turn straight into copy.
Our guide to survey questions covers how to phrase them.
- For the bird food brand, Oliver Kenyon took three reassurance lines from post-purchase survey answers (freshness, aflatoxin testing, delivery to the door) and placed them above the collection grid. The result: $44,890 a month, or $0.66 per session.
- Kasper Dolk saw customer survey objections like "is it durable?" and "is it the official X?" and answered them in benefit bullets in the product hero, for an estimated +$763,679.14 profit per year.
- For a supplement brand, Kasper Dolk turned survey doubts ("what if it doesn't work," "is it a scam," "price is high") into a "how you'll feel" timeline section on the PDP. Conversion rate rose 14%, though Kasper Dolk described it as still being confirmed.
Qualitative data can overturn a best practice
The two countdown timer tests above point in opposite directions, and qualitative research explains why.
Kasper Dolk ran a post-purchase survey for a Dutch audience where customers said the checkout timer felt like too much pressure.
Hiding the timer raised conversion rate 25.61% and revenue per visitor 26.18%.
So urgency worked on one collection page and hurt in one checkout.
Without the survey, you'd have no reason to test removing something that "best practice" says should help.
Usability tests and the five-user rule
Jakob Nielsen's 2000 article on testing with five users argues that five participants uncover about 85% of usability problems.
The model behind it assumes each user finds about 31% of problems, an average across past projects, and Nielsen recommends several small rounds over one large one.
That number is about finding problems, not measuring them.
Five people can show you that a shipping cost surprise confuses shoppers.
They can't tell you what share of your visitors it affects, or whether fixing it will move revenue.
When qualitative research is the wrong tool
- Proving impact. A survey answer is a hypothesis, not a result. Clean Commit made a free-shipping threshold more visible after survey answers named delivery cost as a near deal-breaker. After 27 days the variant showed +2.4% conversion rate with a 74.6% probability to beat control: directional, not significant.
- Ranking problems by size. The loudest complaint in interviews isn't always the most common. Check it against analytics before you build a test around it.
- Asking people to predict behavior. "Would you buy if..." questions are weak evidence. Ask about what already happened instead.
Session recordings: qualitative evidence at quantitative scale
Recordings let you watch individual visitors rage-click, hesitate or scroll back and forth.
Each one is a qualitative observation, but you can watch dozens from the exact funnel step your analytics flagged.
The risk is anecdote. One memorable session where a visitor fought with a size picker can feel like proof.
Before you build a test around it, count how many sessions show the same behavior, or check the matching event in analytics.
Putting the two together
The pattern in the tests above is the same each time:
- Use quantitative data to find the leak (a high exit rate, a weak funnel step).
- Use qualitative research to learn why people leave there.
- Turn the strongest reasons into a hypothesis and copy written in customers' words.
- Run an A/B test and judge it on revenue, not only clicks.
A few practical rules make this loop work:
- Start with quantitative when you have traffic. It stops you spending interview time on pages that barely affect revenue.
- Start with qualitative when you don't. A store with low traffic can still learn a lot from ten customer conversations, even if it can't run a clean A/B test.
- Keep the customer's wording. In the survey-led tests above, the winning copy repeated the objection almost word for word.
- Expect some ideas to lose. Research raises the odds of a good hypothesis. It doesn't guarantee one, which is why the test comes last.
@oliverkenyon
@DaveDiederen
@kasper_dolk@kasper_dolk
Clean CommitClean CommitFor the full set of methods, start at our customer research hub.