Where do retention save-offers and downgrade upsells appear?

The cancellation flow is the concentrated surface: 332 of 23,407 canonical mobile screens are categorized as cancel-subscription screens, where save-offers and downgrade upsells appear.[3] That's nearly 5x the 70 checkout-upsell screens.[3] If you want to intercept churn with an upsell or discount, the cancel flow is where the corpus says that motion lives.

332 of 23,407 canonical mobile screens are cancel-subscription screens — the home of retention save-offers — Lazyweb Research, July 2026.

Lazyweb Research · n=23407 · Published 2026-07-07

cancellationretentionupsellmonetizationmobile

Finding

Canonical-screen counts for the two monetization-adjacent surfaces:[3]

SurfaceCanonical screensRelative size
Cancel subscription332~5x
Checkout upsell701x

Cancellation is a far larger surface than in-app checkout upsell, reflecting how much effort apps put into the last-chance moment — plan downgrades, pause options, and discount save-offers cluster here.

How to apply

Design the cancel flow as an upsell/retention surface, not just an exit: present a save-offer (discount, pause, downgrade to a cheaper tier) before confirming cancellation. Because it's a well-populated, expected surface (332 screens), users tolerate an offer here more than they would an interruptive banner elsewhere. Keep the offer honest and skippable to avoid dark-pattern backlash.

Caveats

Counts are absolute canonical-screen categorizations, not a share of screens (is_paywall NULL on 21,824/23,407) and not conversion.[3] The presence of a cancel screen does not by itself confirm a save-offer is shown — it's where such offers are captured when present.

The numbers

StatComputed from
332 of 23,407cancel_subscription_screens: 332 canonical screens
70 of 23,407checkout_upsell_screens: 70 canonical screens
Methodology. Universe: 23,407 canonical mobile screens. Method: absolute count by screen category, July 2026. Caveat: absolute counts only; is_paywall NULL on 21,824/23,407 rows.

Sources & citations

  1. [3] Lazyweb Research analysis of 23,407 canonical mobile screens (tracked app corpus), July 2026. Canonical-screen category counts; is_paywall is NULL on 21,824/23,407 rows so gating is reported as app-count prevalence or absolute screen counts, never as a share of all screens.

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