What Are Companies A/B Testing In Landing-Page Navigation Right Now?

Across 144 navigation experiments tracked by Lazyweb Research (all web, out of 2,160 area-annotated experiments), the dominant move is stripping the nav down, not adding to it [1]. NAV is the sixth most-tested area but carries the lowest model-assigned impact of any major area (avg 2.74/5), so it reads as a hygiene lever rather than a conversion swing [1]. Teams most often delete escape-hatch links on conversion-critical pages and collapse two-sided navs to a single audience [2].

NAV accounts for 144 of 2,160 annotated experiments (6.9%) and shows the lowest model-assigned impact of any major page area at 2.74/5, based on Lazyweb Research's July 2026 analysis.

Lazyweb Research · n=144 · Published 2026-07-07

landing-pagewebux-patternsexperimentsdesign

The finding: nav gets de-chromed, not expanded

NAV is a web-only area in the corpus: all 144 of its annotations sit on web pages, spread across 135 distinct experiments [1]. It is the sixth most-tested area overall at 6.9% of the 2,160 annotations, but its average model-assigned impact of 2.74/5 is the lowest of any major area, and only 8.3% of nav annotations (12 of 144) were scored high-impact (4+/5) [1][3]. The recurring pattern is removal: on conversion pages, companies cut nav links that give visitors somewhere else to go [2].

How nav sits against other tested areas

NAV is tested far less than the hero and offer areas and scores lower on impact [1].

AreaAnnotations% of 2,160Avg impact (1-5)High-impact share
HERO48823.2%3.2828.1%
NAV1446.9%2.748.3%
SECONDARY ACTIONS1125.3%2.909.8%

Two observed directions dominate: removing the global nav and topic bar on a newsletter/list-building page so the header is just a logo, and collapsing a dual consumer+organizer nav to a single organizer-only nav with four links on a two-sided platform [2].

How to apply it

Treat nav edits as leak-plugging on pages with a single conversion goal, not as a headline growth bet. The impact scores say most nav changes move the needle modestly [1]. The highest-value version is checkout-style de-chroming: on a page whose only job is one conversion, every nav link is a tempting exit [2]. On two-sided products, audit whether the consumer-facing nav is leaking B2B intent out of the funnel [2].

Caveats

All 144 nav annotations are web, and per Lazyweb Research's rules web cuts are reported as absolute counts, not percentage league tables [4]. Impact is a model-assigned 1-5 score on observed before/after diffs, not measured lift, so use it only for relative ranking across areas [4].

The numbers

StatComputed from
144 NAV annotations (6.9% of 2,160), avg impact 2.74/5, avg confidence 3.34, 135 experiments, all webarea_NAV
Observed nav moves: CBS News removed global nav + 11-item topic bar to a lone logo; Eventbrite reduced dual consumer+organizer nav to a single 4-link organizer navqualitative NAV cbs-news, eventbrite
NAV high-impact (4+/5) share 8.3% (12 of 144); HERO 28.1%, SECONDARY ACTIONS 9.8%high_impact_share_by_area
Web corpus is 365 experiments / 880 annotations; impact is a model-assigned 1-5 score on observed diffssmallSampleWarnings; universe
Methodology. Universe is 144 NAV annotations within 2,160 area-level annotations over 1,126 detected web/mobile experiments in the ~800 tracked-apps corpus. Areas are LLM-annotated from observed before/after screenshots; impact is a model-assigned 1-5 score, July 2026, usable only for relative ranking.

Sources & citations

  1. [1] Lazyweb Research analysis of 144 navigation experiments (web landing pages in the ~800 tracked-apps corpus), July 2026. NAV area annotations within 2,160 area-level annotations covering 1,126 detected experiments.
  2. [2] Lazyweb Research analysis of 2,160 annotated UI experiments (~800 tracked apps), July 2026. Named examples are individual observed before/after diffs with LLM-inferred rationale, not measured results.
  3. [3] Lazyweb Research analysis of high-impact-share by area (n>=30 areas), July 2026. Share of annotations scored impact 4+/5 by the model; relative ranking only.
  4. [4] Lazyweb Research analysis of 4,814 detected experiments (~800 tracked apps), July 2026. Methodology caveats: web is thin; impact/confidence are model-assigned 1-5 scores.

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