Qualitative research
Quantitative A/B testing tells you what changed. Qualitative research tells you why. The two complement each other - you need the research to find the friction worth fixing, then the test to confirm that fixing it actually helps.
The main qualitative methods, roughly in order of cost vs depth:
Heuristic reviews
Section titled “Heuristic reviews”A trained CRO practitioner walks through the site as a fresh user, noting friction points against a heuristics checklist (Nielsen’s 10, the usual e-commerce conventions, accessibility basics). Cheap and fast. Catches the obvious issues but doesn’t reveal the non-obvious ones that show up in actual user behaviour.
Worth doing as a baseline before any deeper research. The findings usually populate the first quarter of test backlogs in a new programme.
Session recordings
Section titled “Session recordings”Tools like Hotjar, FullStory, and Microsoft Clarity record real user sessions. You watch playback to see where users hesitate, rage-click, or abandon. Patterns emerge after watching 20-30 sessions of the same flow.
The value is in the unexpected. Users do things you didn’t predict - tap on non-interactive elements, scroll past key content, abandon at unexpected steps. Generates hypotheses you wouldn’t have come up with from heuristics alone.
Surveys
Section titled “Surveys”On-site surveys (post-purchase, exit intent, abandoned cart follow-up) ask users why they did what they did. Useful for the why behind quantitative drop-offs. The trap is self-report bias - people aren’t always good at explaining their own behaviour.
Open-text questions usually beat multiple-choice for qualitative depth. “What almost stopped you from buying today?” produces more useful data than “Rate your experience 1-5”.
User interviews
Section titled “User interviews”Scheduled conversations with actual customers or prospects. Highest signal-to-noise of any qualitative method when done well, highest cost in time and recruiting. Best for understanding the outcome the buyer is after and their broader context, not for tweaking single-page UX.
The “Switch interview” framework - walk a buyer backward through what was happening before, during, and after their purchase - is the canonical CRO-relevant interview structure.
Usability testing
Section titled “Usability testing”Watch a small number of users (5 is usually enough to surface major issues) attempt specific tasks on the site. Moderated or unmoderated. The five-user rule from Nielsen Norman is the standard reference - past 5, the marginal new issue found per user drops sharply.
Mostly useful for catching show-stoppers in new flows before they ship. Less useful for fine optimisation of existing flows.
Where qualitative doesn’t help
Section titled “Where qualitative doesn’t help”Qualitative research is for generating hypotheses, not for validating them. The trap is treating qualitative findings as conclusions. “5 users said the CTA is confusing” isn’t evidence the CTA is hurting conversion - it’s a hypothesis worth testing. Until you A/B test the alternative, you don’t actually know.
The other trap is the loud minority. Survey responses and interview volunteers skew toward people who care enough to engage. The silent majority who quietly converted (or quietly left) is underrepresented. Quantitative measurement is the counterweight.