How Much Do Apps Experiment On Their Cancel-Subscription Flow?

Across 4,814 detected experiments (July 2026), cancel-subscription is the second most-experimented labeled screen category with 371 distinct experiments — trailing only the home screen (562) and ahead of every paywall category. [1] Retention-flow iteration is a first-class testing surface in this corpus, not an afterthought. These are observed before/after variations with inferred rationale, not company-confirmed A/B tests.

Cancel-subscription is the 2nd most-experimented screen category in the corpus at 371 distinct experiments (July 2026).

Lazyweb Research · n=4814 · Published 2026-07-07

cancellationretentionexperimentsmonetizationmobilesaas

The finding

The cancel-subscription flow ranks second among all labeled screen categories at 371 distinct experiments, behind only the home screen (562) and ahead of account-login (167), sign-up (127), and the paywall/subscription category (117). [1] That places the retention/save flow above the acquisition paywall itself in detected testing volume — a signal that top apps treat the cancel path as a serious optimization surface.

Screen categoryDistinct experiments
Home562 [1]
Cancel subscription371 [1]
Account login167 [1]
Sign up127 [1]
Paywall / subscription117 [1]

Why the cancel flow gets so much attention

Every dollar saved in a cancel/save flow is a dollar that did not require re-acquisition, so the ROI on a winning save-offer or friction test is high. That 371 detected experiments sit above the paywall category (117) suggests the most-tracked apps have already learned retention interventions — pause offers, discounts, downgrade paths — are worth continuous testing.

How to apply it

If your subscription app tests the acquisition paywall but ships a static cancel flow, this data says you are under-investing in the corpus's second-most-tested surface. Start with a save-offer test (pause, discount, or plan downgrade) at the cancel step. Treat these as detected patterns worth trialing, not proven winners. [1]

Caveats

The 371 count is a lower bound because screen category is NULL on 1,425 of 4,814 experiments. [2] All are observed variations with LLM-inferred rationale, not company-confirmed A/B tests — no lift is measured, and the cancel-flow category is not attributed to specific named companies in this stat pack. [1]

The numbers

StatComputed from
Cancel-subscription 371; home 562; account-login 167; sign-up 127; paywall/subscription 117top_experiment_screen_categories (home=562; cancel_subscription=371; account_login=167; sign_up=127; paywall_subscription=117)
1,425 of 4,814 experiments have no screen categoryscreen_category_null_on_experiments (1425/4814)
Methodology. Universe: 4,814 detected experiments across 276 companies, July 2026; screen categories from screen_category join. Method: COUNT(DISTINCT experiment_id) by category. Caveat: counts are lower bounds (1,425 unlabeled); detected variations only, never confirmed A/B tests.

Sources & citations

  1. [1] Lazyweb Research analysis of 4,814 detected experiments (276 companies, ~800-app mobile corpus), July 2026. Top labeled screen categories by COUNT(DISTINCT experiment_id); cancel_subscription=371.
  2. [2] Lazyweb Research analysis of 4,814 detected experiments (276 companies, ~800-app mobile corpus), July 2026. screen_category is NULL on 1,425 experiments, so all surface splits are lower bounds.

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