Working · 2026-11SecondaryActivationAnalytics

Find the aha moment by comparing retained vs churned users

Compare early behaviour of users who stuck around with those who left, and pick the action that best separates them.

Claim and evidence

The claimthe average retained user spent over 3 minutes in their first session, while churned users only spent 35 seconds

Groove · appcues.com · undated

What the numbers show

over 3 minutes

average first-session time of retained users (Groove, via Appcues)

appcues.com
35 seconds

average first-session time of churned users (same comparison)

appcues.com

Why it works

Diagram, not data
New users' earlybehaviourRetained: did the keyactionChurned: skipped itPush everyone towardthat action
New users' early behaviourRetained: didthe key actionChurned:skipped itPush everyone toward that action

The action retained users take that churned users do not is a candidate activation milestone, which you can then test by pushing users toward it.

How to run it

  1. Define a retention outcome (e.g. day 30 active)
  2. List early actions for retained and churned cohorts
  3. Rank actions by difference
  4. Run an experiment to push users toward the top action

Ways to run it

Design options to test, not promised results.

  1. Option AFirst-session comparison

    Compare what retained and churned users did in their very first session, as Groove did.

  2. Option BRanked action list

    List early actions for both cohorts and rank them by the gap between the groups.

  3. Option CNudge experiment

    Run a test that pushes new users toward the top action and check whether retention moves.

Dated examples

  1. —

    Groove

    Compared first-session behaviour of retained and churned users.

    Retained users spent over 3 minutes in the first session; churned users 35 seconds.

    “the average retained user spent over 3 minutes in their first session, while churned users only spent 35 seconds”
    ✓ quote checked word-for-word on the page

    appcues.com · undated

Does it still work?

Workingas of 2026-11

Method described in a vendor guide (undated); correlation, not causation, until tested.

Risks

  • Correlation can be selection effect
  • Needs enough users per cohort

Sources

  1. appcues.com · Appcues · Ty Magnin · undated
    “the average retained user spent over 3 minutes in their first session, while churned users only spent 35 seconds”
    ✓ quote checked word-for-word on the page

Change log

  • 2026-11-20 — harvest v0.1 (activation)