What to test
📖 3 min readUpdated 2026-04-19
Not all tests move the needle equally. Prioritize for impact.
High-impact tests
- Headlines
- Hero images
- CTA copy
- Pricing structure
- Form length
- Page structure (radical changes)
Lower-impact
- Button color
- Microcopy
- Font choices
- Minor visual tweaks
Skip
- Things with tiny traffic
- Changes you can't measure
- A/B tests driven by opinion not data
What to do with this
- Test the offer and headline before testing buttons, the 30-100% lift potential is where real gains come from
- Prioritize tests by expected lift × traffic, a 30% lift on low-traffic page may be less valuable than a 10% lift on high-traffic page
- Generate hypotheses from research (analytics + heatmaps + interviews), opinion-driven tests produce noise
- Keep a backlog scored with ICE or PXL, picking tests from a ranked backlog beats picking from intuition
- Don't test changes you can't measure, vague hypotheses produce vague results