How many CTAs should I put in the hero?
- By Ali Abouelatta
- n = 488
- Published
- Updated
- 2 min read
- landing-page · experiments · ux-patterns · web · design
Short answer
An exact CTA-button count across landing pages is not computable from this data — the web landing corpus has no join to the vision-JSON source that records button counts.source[1] What the experiments show is a dual-CTA pattern (primary self-serve + secondary demo) emerging as companies move upmarket, seen in detected hero changes.source[2] Use the qualitative pattern, not a fabricated average.
Exact hero CTA counts aren't computable; the detected pattern is primary-plus-secondary as brands move upmarket — July 2026.
Finding: the count itself isn't measurable here
Per-landing-page hero anatomy — including exact CTA-button count — cannot be computed because sites_screen_labels and the vision-JSON screenshots table do not join.source[1] So there is no honest 'average number of hero CTAs' to quote from this corpus. Beware any benchmark that claims one from this data.
Breakdown: the dual-CTA pattern from experiments
The experiment rows reveal a directional pattern (observed change + inferred rationale):source[2]
- Calendly (web): added dual hero CTAs — 'Sign up for free' (primary) + 'Get a demo' (secondary outline) — where the control had no in-hero CTA. Inferred rationale: pairing self-serve signup with a demo escape hatch as the company shifts toward team/enterprise buyers.
- Factory (web): collapsed a multi-element CLI install widget into a single 'START BUILDING' button. Inferred rationale: enterprise-buyer traffic isn't terminal-ready, so a single account-first action fits better.
Direction depends on buyer: upmarket moves add a secondary demo CTA; simplification collapses to one.
How to apply
Let buyer mix decide count, not a fictional average. If you serve both self-serve and enterprise, the Calendly pattern supports one primary self-serve CTA plus a secondary demo path. If your control over-complicates the hero, the Factory pattern supports collapsing to a single action. Both are observed changes with inferred rationale, not measured lift.source[2]
The numbers
| # | Stat | Computed from |
|---|---|---|
| [1] | Exact per-page hero CTA-button count not computable (no web-to-vision-JSON join) | smallSampleWarnings[web-mobile join] |
| [2] | Calendly dual-CTA add and Factory single-CTA collapse examples | qualitative[calendly, factory] |
Methodology
Universe: 1,238 web landing pages plus named-company hero experiments, July 2026. Method: attempted census (blocked by missing join) plus qualitative experiment rows. Caveat: exact CTA count is not computable; examples are observed changes with inferred rationale.
Sources
- [1]
Lazyweb Research analysis of 1,238 landing pages (web landing census), July 2026
No working join to vision-JSON button counts; exact hero CTA count not computable.
- [2]
Lazyweb Research analysis of 405 named-company hero experiments (detected-experiment corpus), July 2026
Observed before/after hero CTA-structure changes with model-inferred rationale.
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.