Why new users stall between install and activation
The AARRR framework puts activation straight after acquisition: the first moment a new user gets the value they installed for, pinned to an action that clearly separates the people who stay from those who drift away. The activation event tutorial picks that action from the app's own data. Between the install and that action sits a short run of steps, and new users drop out at each of them for ordinary reasons: they were interrupted, the step asked for something they did not have to hand, they did not see why it mattered, or the app asked for an account, a profile or a permission before showing anything useful.
The onboarding guide's advice is to remove as much of that run as possible first, by deferring whatever can wait. What remains is the job of onboarding messages, and the guide gives them a simple shape: a single goal for each message, aimed at the next most useful step, and silence once the person has taken it. The push and in-app messaging guide adds the order of channels: guidance in the app comes first, and push is kept for those who leave before they finish. The difficulty is not the principle but the build. Teams without a step map send the same reminders to everyone on a calendar, which guides people who have already finished and misses people stuck on the hardest step.
| What you see | Where it shows | What it usually means |
|---|---|---|
| New users who finished setting up still get "finish setting up" | Messages sent against the events each recipient had already logged | The sequence runs on days since install, not on what the person has done |
| A person stuck at one step gets the same reminder in the app, on the lock screen and by email on one day | Messages per person per day across tools | Each channel's team built its own onboarding with no shared plan |
| Notification disables climb in the first weeks after install | Disables by days since install | Early pushes are spending the permission on nudges the app could show in-session |
| Activation is reported as better since the messages started, with no comparison group | The onboarding report | No holdout; the messages take credit for people who would have activated anyway |
| Messages still point at a screen the last release removed | Taps that open the home screen or an error | Nobody reviews the sequence when the first session changes |
The diagnosis of installs that do not turn into active users separates a campaign, a listing or a first-session problem from a messaging one. Fix a wall in the first session before building messages around it; a reminder to climb a wall that should not be there only annoys more people.
Before you start
Gather the following, and settle the definitions before anyone writes a message.
- The activation event and its window. One event, written down with its window: how long after install it still counts, shared by product, analytics and whoever reports onboarding. The activation tutorial produces it.
- The first-session screen map. Each screen from first open to the activation event, as the onboarding guide asks for, with the steps that were deferred and when each is now asked.
- Events for every step. An event for each step's completion, logged with the same user identifier the messaging tools use, so a message can be triggered by a step not yet taken and stopped by a step completed. The deep links and app analytics guide sets out the event taxonomy and the audiences built from events and properties.
- The tools that send. The engagement tool for push and in-app messages, and the email platform if new users give an address. Firebase's documentation, as the push guide reports it, covers push through Cloud Messaging and in-session messages through In-App Messaging, each triggered by events; other engagement tools follow a similar model.
- The programme's rules. The shared frequency cap, quiet hours and priority order from the push guide, and the notification topics from the tutorial on notification preferences by topic. The customer contact policy template keeps these rules in one document beside the email and text caps.
- Somebody who can talk to new users. Event data shows where people stop; a few conversations with recent installs who stopped, run as the customer interviews tutorial describes, usually show why.
- An owner for the whole sequence across channels, and a date to review it.
The steps
Map every step between the first open and activation
Write the steps a new user takes from the first open to the activation event, in the order the app presents them: browse, choose, create an account, add the detail the service needs, confirm, complete the core action. Include the steps the onboarding guide deferred, at the point they now appear, and mark which ones a deferred deep linkA link that opens a specific screen inside a mobile app rather than its home screen. skips for people who arrived from an ad or a shared link. Next to each step, write the event that records its completion. A step without an event cannot trigger a message or stop one, so add the missing events to the tracking plan before going further.
Find where each install cohort stalls
Read the steps as a funnel for each weekly install cohortA group of customers who share a starting event and period, tracked together over time.: the share of the cohort that completes each step, and how long the people who finish usually take. The conversion rate calculator gives the rate between any two steps from the counts. Mark the steps where most of the loss happens; the rest need no message. Keep iOS and Android apart, since each asks for permission differently and a step can stall on one platform only, and treat small cohorts as the guide to reading small numbers advises, waiting for a pattern to repeat before building on it. Then ask a few of the people who stalled what stopped them, because a missing detail, a confusing screen and a simple interruption each call for a different message.
Write one job for each stalled step
For each marked step, write in one line what the message has to achieve and for whom: "people who chose a tool but did not add a delivery address", not "drive engagement". Say what completing the step gives the person in terms of what they installed for, and what would make it easier: a shortcut, a saved choice, an answer to the question that stopped them. One job per message is the guide's rule; a message that tries to finish two steps finishes neither. Note whether the job is service or promotional. Help to complete something the person started is usually service content; anything that promotes an offer is promotional, and the notification preferences tutorial treats a message that mixes the two as promotional.
Put the first nudge in the app, where the step lives
Build the first message for each step as an in-app messageA message shown inside an app during a session the person started, such as a banner, modal or tooltip, needing no operating-system permission., shown to people in a session who have not completed the step. The guide to designing in-app messages chooses the format: a card that resumes an unfinished task when the person returns, a tooltip on the screen where the step happens, a banner after a related action. It also rules out the modal on every open. Trigger the message from behaviour, such as reaching the screen or returning with the step unfinished, never from the app opening, and give it the exit condition the guide asks for: it disappears once the step is done.
Add a push only for people who allowed it and left
A push notificationA message an app sends through the operating system's notification service, shown on the device outside the app to people who allowed notifications. reaches only people who allowed notifications, and the push guide reserves onboarding push for those who leave before finishing. Send it after a wait that gives the person a fair chance to come back on their own, set from how long the people who finish usually take, and only if the step is still open. Write it as the push writing tutorial describes, with what is waiting for this person first and the specific thing named, and link it to the unfinished step through a deep link, checked with the deep link test for the signed-out and new-install cases. Under Apple's App Store Review Guidelines, as the push guide reports them, marketing push needs the person's opt-in, so an onboarding push that promotes an offer goes only to people who chose promotional messages.
Agree one channel per nudge with the email flow
If new users also share an email address, the tutorial on software onboarding emails builds the email side and sends an email only when the person leaves without completing the step. Put every onboarding message from every channel in one table: the step, the trigger, the channel, the wait and the exit. For each nudge, decide which channel carries it for which person, as the onboarding guide asks, so nobody who stalls is nudged about one step in the app, on the lock screen and in the inbox on the same day. The table in the next section is a starting point; the routing comes from who can be reached, not from which team built the message first.
Set exits on the step, on activation and on silence
Every message needs three exits. The step: once it is done, its in-app card disappears and its push and email are never sent. Activation: once the person reaches the activation event, every onboarding message on every channel stops and the programme for active users takes over. Silence: a person who has not opened the app since the push is not helped by a second and third reminder, so give each step its push and, for email, at most one follow-up as the email tutorial does, then leave them to the re-engagement guide for lapsed app users, which defines a lapse and the channels that can still reach one. Check exits before each send, not only on entry, because a message queued before the step was completed must not go out after it.
Fit every message under the shared cap, quiet hours and topics
Onboarding messages share the programme's frequencyAverage times a person saw an ad. cap with everything else the app sends, as the push guide sets it: a single cap for each person over each period, shared by every team and campaign. Place onboarding above promotions in the priority order and below transactional messages, such as a booking confirmation, which sit outside the cap. Schedule push in the recipient's local time and hold non-urgent onboarding pushes that fall in quiet hoursThe hours when promotional texts, WhatsApp messages and push notifications are held for the recipient's daytime, set by law in places and courtesy elsewhere. until morning. Give each onboarding push a topic among the app's notification settings, letting someone who wants order updates but not tips say so. The mobile message send QA checklist ticks consent, caps, timing and the link before anything goes live.
Keep a holdout and read activation by install cohort
Keep a random share of new users out of the sequence, receiving only service messages, and leave it in place for several cohorts. A holdout groupA randomly chosen part of an audience kept from exposure to measure what exposure caused. is the only way to see what the messages add, because many people who receive an onboarding nudge would have finished the step anyway. For each weekly install cohort, compare the activation rateThe share of new users or sign-ups who complete the activation event, the action marking first real value, within a set window. within the agreed window between the people who got the sequence and the holdout, and the costs beside it: notification disables, topic opt-outs, uninstalls and messages per person. Read only cohorts whose window has closed, build them as the cohort analysis tutorial describes, and treat a small difference as noise until it repeats. Run changes to a single message as tests inside that structure.
Review the sequence when the first session changes
A release that moves the account step, adds a screen or renames one changes the step map, and a message pointing at the old step is now wrong. Make the sequence's table part of the release review for anything that touches the first session, retest the deep links each message uses, and reread the stall points for the first cohorts after the release. Retire messages for steps that no longer stall. Watch the push opt-in rateShare of app users shown the notification permission prompt who allowed notifications. and the share of new users with notifications still on: if they fall after the sequence starts, the diagnosis of users turning off notifications checks the cap, the timing and the topics before the messages are blamed or kept.
Which channel carries which nudge
The routing below is general practice built from the push guide's order, the onboarding guide's single decision per nudge and the email tutorial's rule that email follows the person out of the product. It decides by the person's state, not by the message.
| The new user's state | In-app | Push | Stops when | |
|---|---|---|---|---|
| In a session, step not done | The first nudge: a card, tooltip or banner where the step lives | No; they are already in the app | No | The step is done |
| Left the app, notifications allowed | Waits for their return | One deep-linked push after the wait | Only if the plan gives email this step | The step is done, or they go quiet |
| Left the app, notifications not allowed or not yet asked | Waits for their return | Not available | One email if they gave an address and the step is in the email flow | The step is done, or they go quiet |
| Reached activation | Messages for active users only | Onboarding stops | Onboarding stops | Immediately, on every channel |
| Silent since the last nudge | Shown if they return | No more onboarding pushes | No more onboarding emails | Handed to the lapsed-user plan |
Two rows explain most mistakes. The first shows why onboarding push sent to people already in a session is waste: the in-app message can say the same thing at the moment it is useful, without spending the permission. The third shows why asking for notifications early does not rescue onboarding: people who have not allowed notifications can still be guided in the app on their next visit and, where they gave an address, by email.
A tool hire app builds its onboarding messages (illustrative)
Fexhorvin Tool Hire replaces its reminders with a step-triggered sequence
Fexhorvin Tool Hire, an invented app for hiring power tools by the day, defines activation as a first completed hire within 14 days of install. It had been sending every new user a push on days 1, 3 and 7 saying "Ready to start your next project?", whatever they had done. The step map shows four steps to a first hire: choose a tool, pick a collection or delivery slot, add an address and payment card, confirm. Weekly install cohorts of about 2,400 people lose most people at two of them: 46% of new users choose a tool but never pick a slot, and 31% of those who pick a slot stop at the address and card screen. Interviews with a dozen people who stalled show that many were checking whether the tool would arrive in time for the weekend, and that the card screen asked for details before showing the total.
The team builds two in-app cards, one that reopens the chosen tool with the earliest slot shown, and one that shows the full price before the card form. People who leave without picking a slot and allowed notifications, about 38% of installs, get one push the next evening, "Your hedge trimmer can be ready by Saturday. Pick a slot", deep-linked to the slot screen. Address-and-card stallers who gave an email address get one email instead of a push. The old day-numbered pushes stop. A tenth of each cohort is held out and gets only booking confirmations.
| Install cohorts | Group | New users | First hire within 14 days | Activation rate | Onboarding messages per new user | Notification disables in first 14 days |
|---|---|---|---|---|---|---|
| Weeks 1 to 4 | Sequence | 8,600 | 1,720 | 20.0% | 1.4 | 3.1% |
| Weeks 1 to 4 | Holdout | 960 | 158 | 16.5% | 0 | 2.9% |
| Weeks 5 to 8 | Sequence | 8,700 | 1,775 | 20.4% | 1.3 | 3.0% |
| Weeks 5 to 8 | Holdout | 970 | 162 | 16.7% | 0 | 3.0% |
Across eight weekly cohorts the sequence activates 20.2% against 16.6% for the holdout, with disables close to the holdout's, where the old day-numbered pushes had run at 5.8%. Fexhorvin keeps the sequence and the holdout, and adds the slot card to the release review so the next redesign of the booking screens updates it.
Common mistakes
- Messaging around a wall. A reminder to complete an account form that should not be in the first session only reaches more people with the same obstacle. Defer what can wait first, then build messages for the steps that remain.
- A calendar instead of a step map. Messages on fixed days after install guide people who have already finished and miss people stuck on the hardest step. Trigger each message from a step not yet taken.
- Push before in-app. A person still in the session can be helped on the screen in front of them. Keep push for people who allowed notifications and then went away, and send it once per step.
- The same nudge on every channel. Separate teams building onboarding in the engagement tool and the email platform produce three reminders a day. Keep one table and decide one channel per nudge.
- Exits checked only on entry. A message queued before the step was done goes out after it. Check the step and the activation event before every send.
- Onboarding outside the cap. Every early push spends a permission the app has only just been given. Count onboarding against the shared cap and keep promotions in it below the onboarding messages.
- Crediting the sequence with every activation. Without a holdout, the messages take credit for people who would have finished anyway. Keep a holdout and read activation by install cohort once the window has closed.
- A sequence nobody updates. A release that renames a step leaves messages pointing at nothing. Review the table and retest its links whenever the first session changes.
