Retention Agent

Find the cohortthat's churning —and fix it.

The Retention Agent monitors subscription churn cohorts across trial source, geography, plan, and onboarding flow. It surfaces the single cohort with the most addressable churn this week and prescribes a specific fix — a win-back trigger, an onboarding tweak, or a plan change.

What it does

The Retention Agent slices your subscription base by cohort attributes — trial source, country, plan, app version, and onboarding flow — and identifies which cohort has the most addressable churn relative to its size.

It distinguishes between unavoidable churn ("the user got what they wanted") and addressable churn ("the user hit a paywall friction point on day 7") so you spend your week on fixes that actually change behavior.

It also tracks the impact of fixes you ship over time, so a retention tweak that worked for one cohort is automatically extended to similar cohorts.

How it works

  1. 01

    Pull subscription cohorts

    Reads active subscriptions, churn events, and refund data from RevenueCat by cohort attributes.

  2. 02

    Score addressable churn

    Distinguishes unavoidable churn (got value, churned) from addressable churn (hit friction, churned) using paywall and engagement signals.

  3. 03

    Propose a single fix

    Surfaces the cohort with the most addressable churn this week and proposes a specific fix — win-back, onboarding tweak, or plan adjustment.

  4. 04

    Track impact

    Monitors retention curves on the affected cohort post-fix and reports win/loss in your weekly plan.

Sample weekly recommendation

"Annual-trial users from TikTok ads who churn in week 2 (n=412 last 30 days) are not opening the app between days 4 and 9. Ship a day-5 in-app message featuring your highest-rated feature ("AI photo enhancer"). Expected lift: 8-14% week-2 retention on this cohort."

Integrations and pricing tier

Integrations used

  • RevenueCat
  • App Store Connect

Tier required

Available on all tiers. Win-back campaign execution on Growth and Studio.

Frequently asked questions

Cohorts are defined by acquisition source, country, plan type, app version at signup, and onboarding flow path. The agent automatically clusters similar cohorts so you do not have to define segments manually.

Most often: in-app win-back triggers (push notification, in-app message), onboarding tweaks (move a feature higher in the flow), and plan changes (offer a downgrade-instead-of-cancel path). The agent picks based on what historically moved the needle for similar cohorts.

Today the agent surfaces recommended copy and timing — you implement the trigger in your notification provider. We are evaluating direct integrations with OneSignal and Customer.io.

It uses engagement signals (sessions, key actions completed) and paywall friction events to distinguish "got value, churned anyway" from "hit a friction point, churned." Only addressable churn drives recommendations.

Cohort-specific retention curves typically need 2-4 weeks to stabilize statistically. The agent surfaces preliminary signal as early as week 1 with explicit confidence ranges.