Should I test the hero headline or the hero media first?
Test the headline first: headline rewrites appear in 277 of 488 detected hero experiments (57%), versus 68 (14%) for video/cinematic media and 42 (9%) for product-screenshot media.[1] Copy is where teams concentrate hero tests by a wide margin. Note these are keyword-incidence counts over experiment text, not a census of live landing pages.
Headline rewrites appear in 57% of hero tests vs 14% for video and 9% for screenshots — July 2026.
Finding: copy dominates media in hero tests
Among 488 hero experiments, text changes vastly outnumber media changes.[1]
| Hero lever | Experiments | Share of 488 |
|---|---|---|
| Headline rewrite | 277 | 57% |
| Video / cinematic media | 68 | 14% |
| Product screenshot / mockup | 42 | 9% |
Headline work is roughly four times more common than any single media swap.
Breakdown: what a headline test typically does
Detected headline changes cluster around repositioning, not word-polish (observed change + inferred rationale):[2]
- Calendly (web): 'Book meetings that work for you' became an eyebrow 'CALENDLY FEATURES' + 'More than a scheduling link' with a platform-framing subhead — naming and negating a commoditized perception.
- Framer (web): 'Build better sites, faster' became 'The internet is your canvas' — swapping a speed benefit for an aspirational identity metaphor.
- n8n (web): 'AI agents and workflows you can see and control' became 'Flexible AI workflow automation for technical teams' — naming the audience to self-select technical visitors.
How to apply, honestly
Sequence headline tests ahead of media rework — that is where 57% of hero test energy goes and where the qualitative rows show real positioning shifts. Important caveat: these counts describe what changed in tested experiments, not the media mix of all landing pages. Because the web landing corpus and the mobile screenshot corpus have no working join, per-page hero anatomy (exact media type, CTA count) is not computable here — so this ranks levers, it does not tell you what share of pages use each.[1]
The numbers
| Stat | Computed from |
|---|---|
| Headline rewrite 277 (57%), video 68 (14%), screenshot 42 (9%) of 488 hero experiments | hero_change_themes |
| Calendly, Framer, n8n hero headline examples | qualitative[calendly, framer, n8n] |
Sources & citations
- [1] Lazyweb Research analysis of 488 hero experiments (detected-experiment corpus), July 2026. Keyword incidence over summary+learning text; not a per-landing-page media census (no web-mobile join). ↩
- [2] Lazyweb Research analysis of 405 named-company hero experiments (detected-experiment corpus), July 2026. Observed before/after headline changes with model-inferred rationale. ↩
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.