A user hasn't opened your app in three weeks. Your system flags them churned. A win-back push goes out. According to MWM's win-back campaign research, that message recovers 1 to 5% of the audience it reaches.

The same message sent while the user was still installed recovers 5 to 15%. The gap between those two numbers is timing, and nothing else.

Airbridge's churn research describes users moving from Active to Drifting to At-Risk before they cancel or uninstall. Each stage shows up in behavioural data if you're tracking for it.

FullStory's 2024 analysis of behavioural churn signals makes the same point from the friction side. Churn starts as friction that accumulates over weeks, well before a support ticket or a cancellation. An inactivity trigger reads the outcome of that friction. It never sees the friction.

Session frequency decline is the one most teams eventually notice. Session depth reduction usually shows up first.

When users begin to lose interest, the first sign is that they do less during each session, not that they visit less often. Someone who used to complete four actions may now do only one, even though they still open the app regularly. If you catch this change, you can respond before their usage starts to decline.

The baseline has to be individual. A user who opens a banking app twice a week and a user who opens a delivery app daily cannot share a threshold without generating false positives on one and false negatives on the other.

In-app nudges pull click-through rates of 15 to 40% per Nvecta's 2025 benchmark data, against 2 to 5% for email and push. So when a user starts swiping away in-app content they used to engage with, that's disengagement happening inside a session they chose to be in.

Before you treat rising dismissals as churn risk, check your own send volume for that user over the same window. If volume climbed and dismissals climbed with it, you built the problem. Courier's research found 52% of users who disable push eventually churn from the app entirely, and a share of that is self-inflicted.

Assign points per signal. Sum them per user. Set threshold tiers that fire different campaign types, not louder versions of the same one.

Gartner's 2025 Customer Success research found teams running automated churn early-warning systems cut annual churn by roughly 3.1 percentage points against manual monitoring, and the operational difference was that scores were wired into playbooks rather than sitting in a dashboard. A score nobody acts on inside the two-week window is an observation, and observations don't retain anyone.

The full article breaks down all 6 signals with their detection thresholds, the starting point weights for each one, the 4 risk tiers, and the specific campaign format each tier calls for, including why the discount belongs at touchpoint three and never touchpoint one.

Your MoEngage or CleverTap instance knows a user is a dormant premium subscriber. Your app shows them the same home screen as everyone else. That distance between what your CEP knows and what your app displays is the gap this tool scores.

25 questions, 6 capability areas, roughly 5 minutes. You get a 0 to 100 maturity score, a per-category breakdown, and your top 5 gaps ranked by impact. No signup, and the scoring runs in your browser so nothing hits a server.

→ Score your in-app gap (free, 5 min) CEP In-App Gap Analyzer