CRO Framework: Research, Hypothesis, Test, Learn

Phase 1: Research (Quantitative + Qualitative)

• Quantitative research (the WHAT): Use analytics to identify WHERE visitors are dropping off — which pages have high exit rates, where in the funnel (adding to cart, checkout, payment) drop-off is highest, which devices have lowest conversion rates, which traffic sources convert worst
• Tools for quantitative research: Google Analytics 4 (funnel analysis, landing page performance, user journey reports), Hotjar (session recordings and heatmaps showing WHAT users do), Search Console (which queries drive low-converting traffic)
• Qualitative research (the WHY): Understand WHY visitors drop off — surveys (exit-intent: 'What stopped you from completing?'), user testing videos (watch real people use your site), live chat analysis (what questions do people ask before buying?), customer reviews of competitors (what do customers wish was different?)
• The research phase must answer: What is the primary barrier to conversion on this specific page for this specific type of visitor?

Phase 2: Hypothesis Formation

• A CRO hypothesis has three components: 1) Because [evidence from research], 2) We believe that [proposed change], 3) Will result in [expected outcome] for [visitor segment].
• Example: 'Because 60% of mobile visitors drop off on the payment page (evidence), we believe that adding a 'Pay by PayPal' button (change) will reduce payment friction and increase mobile checkout completion by 15% (expected outcome) for mobile visitors (segment).'
• Prioritization: Use the ICE framework — Impact (1-10: how much can this improve conversion?), Confidence (1-10: how strong is the evidence?), Ease (1-10: how easy is this to test?). Test highest ICE scores first.
• Good hypotheses are specific, measurable, and based on evidence. Bad hypotheses start with 'I think...' without any research backing.

Phase 3: Test Design & Execution

• Control vs Variant: Control = current version. Variant = the change you're testing. Run both simultaneously.
• Traffic split: 50/50 split between control and variant for most tests. More variants = more traffic needed for statistical significance.
• Sample size calculation: Use an A/B test calculator to determine how many visitors you need per variant before you can trust the result. Ending a test too early is the #1 CRO mistake — you'll see random noise, not real signal.
• Testing duration: Run for minimum 1-2 full business cycles (typically 2 weeks minimum) to account for day-of-week behavioral variation.
• One change at a time: In a simple A/B test, change exactly ONE element. If you change the headline AND the CTA AND the image simultaneously, you don't know which change caused the result.

Phase 4: Analysis & Learning

• Statistical significance: Your test result is valid only when you reach 95% statistical confidence (standard benchmark). This means there's less than a 5% chance your result is due to random variation.
• If variant wins: Implement the change permanently. Document WHY it won (what does this tell you about visitor psychology?). Design the next test based on these learnings.
• If control wins: The hypothesis was wrong. Document what you learned (negative results are valuable). Form a new hypothesis.
• If inconclusive: The change had no meaningful impact. Move on — don't waste time retesting. Focus testing budget on higher-impact hypotheses.
• The CRO growth loop: Each test generates insights about visitor behavior that inform better hypotheses → better tests → bigger wins. Teams running 5+ tests per month see compounding results. Teams running 1 test per quarter plateau.

Tip

Practice Task

Quick Quiz

Common Mistake