What Are Companies A/B Testing on the Offer (Discounts and Trials)?

Offer — discounts, trials, and deal framing — is the highest-impact high-volume area Lazyweb Research tracks: 301 of 2,160 annotations (14.3%) with an average model impact of 3.69/5, the top score among areas with n≥100. Two-thirds (66.4%) of offer tests are scored 4+/5. It is overwhelmingly a mobile paywall surface — 255 of 301 annotations (85%) are mobile. Observed moves cluster on deepening discounts and making free trials explicit. [1]

Offer is the highest-impact high-volume test area at 3.69/5 average impact, with 66.4% of tests scored 4+/5 across 301 annotations — Lazyweb Research, July 2026.

Lazyweb Research · n=301 · Published 2026-07-07

paywallmonetizationtrialspricingexperimentsmobile

How much, and where

Offer testing at a glance: [2]

MetricValue
Annotations301 (14.3% of 2,160)
Distinct experiments281
Mobile / web split255 / 46
Avg impact (1–5)3.69 (highest of n≥100 areas)
Share scored 4+/566.4% (200/301)

At 85% mobile, offer is a paywall-native area — teams iterate on the deal far more on in-app subscription screens than on web. [3]

Observed patterns

From offer experiments Lazyweb Research scored impact 5: [4]

  • Deepen the discount, hold the framing. Blinkist strengthened its paywall from 50% OFF ($49.99/yr) to 75% OFF ($24.99/yr) while keeping identical 'ONLY TODAY' urgency and CTA — a clean read on discount depth.
  • Make the trial explicit. Audible introduced a free trial: a '1 MONTH FREE' badge, price reframed as '$15.99/mo after trial,' and a CTA changed to 'Sign up for Premium Plus trial' (the control literally said 'There is no trial period').

The recurring bet: reframe the same product as a lower-risk or better-value first step, either by cutting the effective price or by leading with 'free.' [4]

How to apply it and caveats

If you run a subscription paywall, the field says the deal itself is your highest-leverage test surface. The cleanest experiments change one variable — discount depth or trial presence — and hold urgency/CTA constant, so you can attribute the result. Caveat: impact is a model-assigned 1–5 score on observed diffs, a relative ranking, not measured lift. [5] Deeper discounts trade ARPU for conversion, so they fit products whose LTV can absorb the hit — the Blinkist example is an observation, not a recommendation.

The numbers

StatComputed from
Offer 301 annotations (14.3%), avg impact 3.69, 255 mobile / 46 web, 66.4% scored 4+statpack area_OFFER + high_impact_share_by_area
Offer: 301 annotations, 281 experiments, avg impact 3.69, 66.4% scored 4+ (200/301)statpack area_OFFER + high_impact_share_by_area
Offer is 85% mobile (255/301)statpack area_OFFER + platform_area_split
Offer impact-5 examples: Blinkist (50%→75% off), Audible (added free trial)statpack qualitative OFFER entries
Impact is a model-assigned 1–5 score on observed diffs, not measured liftstatpack methodology note
Methodology. Universe: 301 offer annotations across 281 detected experiments (subset of 2,160), July 2026, LLM-labelled and impact-scored from paired paywall screenshots. Named companies are single observations; impact is relative signal, not measured conversion lift.

Sources & citations

  1. [1] Lazyweb Research analysis of 301 offer annotations (281 experiments) within 2,160 total, July 2026. OFFER: highest avg impact (3.69) among n≥100 areas; 255 mobile / 46 web.
  2. [2] Lazyweb Research analysis of 301 offer annotations, July 2026. 200 of 301 scored 4+/5.
  3. [3] Lazyweb Research analysis of 2,160 area annotations, July 2026. Offer is 85% mobile — a paywall-native surface.
  4. [4] Lazyweb Research qualitative review of top-impact offer experiments, July 2026. Blinkist and Audible — single observations scored impact 5 by the model.
  5. [5] Lazyweb Research methodology note, July 2026. Impact is a relative model score on before/after diffs, not measured lift.

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