TL;DR
- Five methods cover 90% of useful CRO research.
- Start with analytics + session replay; escalate to survey and testing as the question sharpens.
- Stop funding research you cannot tie to a decision.
- Quant and qual used together beat either on its own.
Method 1 - Analytics funnel analysis
Step-level drop-offs, device cohorts, audience cohorts. Cheap, fast, often enough to find the first 3-5 hypotheses.
Method 2 - Session replay (sampled)
Watch the sessions that drop off at the problem step. Not all sessions - the ones where the behaviour is most anomalous. Twenty good replays are worth more than a thousand random ones.
Method 3 - Exit and intercept surveys
Three-question micro-surveys triggered on specific behaviours. Short. Specific. Closed-loop with a single open text field. Useful; never the sole basis of a decision.
Method 4 - Heuristic audit
A structured expert review against a documented checklist (not one person's gut). Gives you the long list of candidate issues for research to prioritise.
Method 5 - Usability testing (5-7 participants)
Moderated or unmoderated, task-based. Five to seven is where diminishing returns start. Cheap against the quality of insight.
What we retired
Heatmaps in isolation, scroll-depth vanity, "average time on page" decisions, exhaustive cognitive walkthroughs. None of these changed enough decisions to justify the hours spent.
Sequencing
Analytics first, session replay on the anomalies, heuristic audit to pressure-test, surveys and usability for the sharpest remaining questions. Stop when you have three testable hypotheses - do not research in search of perfection.
Need this installed?
Npexpos's conversion rate optimisation engagement builds the research discipline, test backlog and programme governance in 8 phases.