Can I trust a benchmark that compares the same product's mobile app vs its website?

Only for a narrow set of products, and even then with caveats. Just 168 companies appear in both the 807-app mobile corpus and the 670-site web corpus (21% of mobile apps) [1]. And because all 365 web experiments share one placeholder screen id, web experiments aren't tied to a screen graph at all, so even matched products can't be compared experiment-by-experiment [2]. Trust corpus-level pattern contrasts; be skeptical of same-product claims unless the product is named and confirmed in both.

Only 168 companies (21% of mobile apps) appear in both the mobile and web corpora — Lazyweb Research, July 2026.

Lazyweb Research · n=807 · Published 2026-07-07

experimentsmobilewebsaasdesign

Finding: overlap is small and web experiments aren't screen-anchored

Two limits make most same-product comparisons unreliable. First, the corpora overlap on only 168 companies out of 807 mobile and 670 web [1]. Second, all 365 web experiments point to a single placeholder canonical screen id, while mobile experiments span 1,357 distinct screens — so web experiments are page diffs that can't be joined to the screen graph [2].

CheckValue
Mobile companies807
Web companies670
In both corpora168
Distinct web experiment screens1
Distinct mobile experiment screens1,357

How to apply this

When you see a 'this app's mobile paywall vs its web pricing page' claim, ask two questions: is the product one of the 168 in both corpora, and is it named explicitly? If not, the comparison is really a corpus-level pattern contrast dressed up as a matched comparison [1]. For web, never accept a claim about where on the page an experiment sat — web experiments aren't screen-anchored [2]. Use the honest framing: mobile and web patterns, not the same product's two surfaces.

Caveats

The 168-company overlap is the ceiling on same-product comparisons, not a guarantee any given product is usable [1]. Web experiment positions are simply unavailable from this data [2]. Any page claiming a matched mobile-vs-web result without naming the product and confirming dual-corpus presence should be treated as corpus-level, not product-level.

The numbers

StatComputed from
168 companies in both corpora (807 mobile, 670 web)cross_corpus_company_overlap
Web experiments span 1 canonical screen (placeholder 23588) vs 1,357 for mobileweb_experiments_not_screen_anchored
Methodology. Universe: mobile (807 apps) and web (670 sites) corpora and their detected experiments, July 2026. Method: count company overlap and distinct experiment screens per platform. Caveat: overlap is a ceiling; web experiments are not screen-anchored.

Sources & citations

  1. [1] Lazyweb Research analysis of 807 mobile apps and 670 web sites (cross-corpus overlap), July 2026. 168 companies appear in both corpora.
  2. [2] Lazyweb Research analysis of 4,814 detected experiments (mobile + web corpora), July 2026. All 365 web experiments share one placeholder canonical screen; mobile span 1,357.

Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.

Related questions

Explore the underlying screens, flows, and A/B tests inside Lazyweb. More research