How big is the landing-page benchmark I'm reading these stats against?
- By Ali Abouelatta
- n = 1,238
- Published
- Updated
- 2 min read
- landing-page · web · saas · experiments · ux-patterns
Short answer
The web landing census covers 1,238 desktop landing pages across 347 companies (about 3.6 captures per company), making landing the 5th most common of 15 web screen types.source[1]source[2] Hero and CTA patterns draw from a separate corpus of 488 hero and 232 CTA detected experiments.source[3] Knowing the denominator keeps every downstream stat honest.
1,238 landing pages across 347 companies, plus 488 hero + 232 CTA experiments — July 2026.
Bar chart · Size
Size — Finding: two corpora, never mixed
One block is about 17.4 Size
- Web landing census
- 1,238 pages / 347 companies
- Detected hero experiments
- 488 (405 named / 172 companies)
- Detected CTA experiments
- 232 (225 named / 118 companies)
Finding: two corpora, never mixed
The benchmark has two distinct sources.source[1]source[3]
| Corpus | Size | What it measures |
|---|---|---|
| Web landing census | 1,238 pages / 347 companies | Which pages are labeled 'landing' |
| Detected hero experiments | 488 (405 named / 172 companies) | Before/after hero UI diffs |
| Detected CTA experiments | 232 (225 named / 118 companies) | Before/after CTA UI diffs |
The census tells you scale; the experiment corpus tells you what changed. They do not share a join, so per-page hero anatomy is not computable from the census.
Breakdown: where landing sits among screen types
Landing is the 5th most common of 15 web screen types.source[2]
| Screen type | Pages |
|---|---|
| editor | 2,119 |
| inbox | 1,455 |
| settings | 1,351 |
| feed | 1,252 |
| landing | 1,238 |
Landing pages are captured at meaningful density — enough to rank company-level category cuts where sample allows.
How to apply
When a stat cites '488 hero experiments,' it is describing tested changes, not all 1,238 landing pages. When a stat cites the 1,238-page census, it describes labeled pages, not measured conversion. Keep the two straight and you will read every benchmark on this topic correctly.
The numbers
| # | Stat | Computed from |
|---|---|---|
| [1] | 1,238 landing pages across 347 companies (avg 3.6 captures/company) | landing_labels_total / landing_companies_total |
| [2] | Landing ranks 5th of 15 web screen types (editor 2,119, inbox 1,455, settings 1,351, feed 1,252, landing 1,238) | landing_rank_among_web_screens |
| [3] | 488 hero experiments (405 named / 172 companies); 232 CTA experiments (225 named / 118 companies) | hero_experiments_total / hero_experiments_named_companies / cta_experiments_total / cta_experiments_named_companies |
Methodology
Universe: 1,238 web landing pages (347 companies) plus 488 hero / 232 CTA detected experiments, July 2026. Method: label counts and annotation counts across two non-joinable corpora. Caveat: census and experiment corpora are never mixed; per-page hero anatomy is not computable.
Sources
- [1]
Lazyweb Research analysis of 1,238 landing pages (web landing census, 347 companies), July 2026
screen_type='landing' in sites_screen_labels; 1,238 of 11,753 labeled web screens.
- [2]
Lazyweb Research analysis of 11,753 labeled web screens (web screen census), July 2026
Screen-type counts; landing ranks 5th of 15.
- [3]
Lazyweb Research analysis of 720 hero+CTA experiments (detected-experiment corpus), July 2026
488 hero and 232 CTA annotations; named-company resolution via control_screenshot_id.
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.