What do companies test on mobile vs web, and where does each platform concentrate experiments?
- By Ali Abouelatta
- n = 4,814
- Published
- Updated
- 2 min read
- experiments · mobile · web · pricing · landing-page
Short answer
The experiment corpus is mobile-dominated ~12:1 — 4,449 detected mobile experiments vs 365 web (of 4,814 total) source[1]. Where they test differs sharply: web concentrates on navigation, announcement bars, pricing tables and forms, while mobile concentrates on offers and pricing on the paywall source[2]. Hero and CTA are heavily tested on both. Use the platform-specific hot spots to prioritize your own test roadmap.
4,449 mobile experiments vs 365 web (of 4,814 detected) — a ~12:1 split — Lazyweb Research, July 2026.
Bar chart · Web
Web — Finding: same volume gap, different focus areas
One block is about 9.7 Web
- HERO
- 272
- NAV
- 144
- CTA
- 91
- ANNOUNCEMENT BAR
- 63
- PRICING TABLE / PRICING
- 47
- OFFER
- 46
- FORM
- 37
- SOCIAL PROOF
- 34
Finding: same volume gap, different focus areas
Mobile carries ~12x the detected experiments of web source[1]. The area mix (annotated observations) shows each platform's obsessions:
| Area | Web | Mobile |
|---|---|---|
| HERO | 272 | 216 |
| NAV | 144 | — |
| CTA | 91 | 141 |
| ANNOUNCEMENT BAR | 63 | — |
| PRICING TABLE / PRICING | 47 | 254 |
| OFFER | 46 | 255 |
| FORM | 37 | — |
| SOCIAL PROOF | 34 | 14 |
Web-only areas (nav, announcement bar, form) reflect a page-diff surface; mobile-heavy areas (offer 255, pricing 254) reflect paywall economics source[2].
How to apply this
If you run a mobile app, the corpus says the highest-leverage tests cluster on the paywall offer and price (255 offer + 254 pricing annotations) — test intro pricing, trial framing, and price points before chrome source[2]. If you run a web product, tests cluster on hero messaging, navigation, and pricing-table structure, plus announcement bars and forms that have no mobile-paywall analog. Match your roadmap to where the platform's leverage actually sits rather than copying the other side.
Caveats
Web experiments (n=365) are ~12x fewer than mobile, so every web-side number here is an absolute count with explicit n, never a rate implying parity source[1]. All 365 web experiments share one placeholder canonical screen id, so they are page diffs and not joinable to a screen graph — do not compute web experiment positions source[3]. 'Detected experiment' means an observed before/after change, not a measured A/B lift.
The numbers
| # | Stat | Computed from |
|---|---|---|
| [1] | 4,449 mobile vs 365 web detected experiments (of 4,814) | experiment_platform_split |
| [2] | Area mix: web HERO 272/NAV 144/CTA 91/PRICING TABLE 47/OFFER 46; mobile OFFER 255/PRICING 254/HERO 216/CTA 141 | experiment_area_mix_by_platform |
| [3] | 365 web experiments all share placeholder canonical screen id 23588; mobile span 1,357 distinct screens | web_experiments_not_screen_anchored |
Methodology
Universe: 4,814 detected experiments and 2,160 area annotations across mobile and web corpora, July 2026. Method: count experiments and area annotations by platform. Caveat: detected before/after changes with model-assigned areas, not measured A/B lift; web n is thin.
Sources
- [1]
Lazyweb Research analysis of 4,814 detected experiments (mobile + web corpora), July 2026
4,449 mobile, 365 web.
- [2]
Lazyweb Research analysis of 2,160 experiment area annotations (mobile + web corpora), July 2026
Area mix by platform from experiment annotations.
- [3]
Lazyweb Research analysis of 365 web detected experiments (web corpus), July 2026
All web experiments point to one placeholder canonical screen; captured as page diffs.
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.