What the first session has to achieve
In the AARRR model, the first session is where acquisition hands over to activation. The person has installed the app because an ad, a listing, a friend or a link promised something, and the first session has one job: deliver that promise, or the first real piece of it, before the person's patience runs out. Everything else the app wants from them, an account, a profile, permissions, preferences, is a cost they pay on the way, and each one is a point where some of them stop.
That makes the first session a funnelModel of the customer journey stages. of its own, with the install at the top and the activation moment at the bottom: the first action after which people are markedly more likely to keep using the app, such as a first booking, a first saved item or a first completed lesson. The app's activation moment should come from its own data rather than from a borrowed rule: the tutorial on finding your app's activation event chooses it from candidate actions and retention by cohort, and the mobile measurement hub explains why the measurement plan should name one activation event for the app.
Three outcomes make a good first session. The person reaches the activation moment, or gets visibly closer to it. The permissions they grant are the ones the app needs, granted at a moment they understood why. And they leave with a reason to come back, which later messages can build on. A first session that collects every permission and an account but never reaches the activation moment has optimised the wrong things.
The shortest path to first value
Write down every screen between the first open and the activation moment, then ask of each one what would happen if it moved later or disappeared. The sign-up flows guide applies the same discipline to sign-up on the web: ask only what the next step needs, move everything else to the moment it is used, and keep any setup question that changes the first screen, because its answer makes the first session more useful. In an app the usual candidates for deferring are the same, with a few of their own.
| Usually deferred | Ask instead | Why it can wait |
|---|---|---|
| Account creation | When the person has something to save, book or buy | Browsing and choosing rarely need an account; committing does |
| Profile details | When a feature uses them | Details that change nothing on screen are a cost with no return |
| Feature tour | Never as a sequence; tips at the screen where each feature lives | Tours are dismissed before they are read |
| Notification permission | At a moment notifications are plainly useful | The system prompt is a single chance on iOS |
| Preferences and settings | After first value, with sensible defaults meanwhile | Defaults serve most people; choices can come later |
| Ratings and reviews | After a completed success, much later | The ASO guide puts the rating prompt after a moment of completed value |
New users who arrive from a link can skip further. A deep linkA link that opens a specific screen inside a mobile app rather than its home screen. that survives the install, the deferred deep link described in the deep links and app analytics guide, lets a person who tapped an ad for a particular item open the app on that item, and the guide notes that onboarding can then be shortened because the context is already known. Design the default first run for people whose destination is not recovered, because the guide is clear that matching is limited by platform rules.
Which permissions to ask for, and when
An app may need several operating-system permissions, and each prompt interrupts the person and asks them to trust an app they met minutes ago. The general practice, labelled as such here, is to ask for each permission at the moment the feature that needs it is first used, not in a block at launch, and to ask for nothing the first session does not need.
Notifications. The push guide explains that on iOS an app must ask before sending notifications and the system prompt can be shown only once; if the person declines, the only way back is through the device settings. Recent Android versions require a runtime permission for notifications too. Apple's App Store Review Guidelines say push notifications must not be required for an app to function and must not be used for marketing unless the person has opted in to receive it. The share of people asked who allow notifications, the push opt-in rate, caps every push message that follows, so the moment of asking matters more than the wording of anything sent later.
Tracking, where the app asks. The app attribution guide explains that Apple's App Tracking Transparency framework requires an app to ask permission before linking a person's data with data from other companies' apps and websites for advertising, and what attribution looks like when the person declines, through Apple's SKAdNetwork. Whether to ask is a decision for the people who own measurement and privacy. This guide makes no recommendation about how that prompt is worded, whether anything is shown before it, or what may be offered around it; read Apple's current guidelines and take advice before designing it.
Device features. Any other permission an app uses belongs to the feature that uses it. As general practice, ask when the person first reaches for that feature, so the reason is obvious, and design what the app does if they say no, so a refusal never blocks the core of the app. This guide does not set out the rules for individual device permissions, which the operating systems document.
Soft prompts that tell the truth
A soft prompt is an ordinary in-app screen or card, shown before the system prompt, that explains what a permission will do for this person and asks whether they want it. The push guide recommends the pattern for notifications: if the person agrees, the app shows the system prompt; if they decline, the system prompt has not been used and can be offered again at a later moment of value. The pattern protects a single chance by spending it only on people who have already said yes.
A soft prompt earns that trust only if it is honest. It should name what will be sent, in the person's terms ("we'll tell you when your walker is on the way"), not the app's ("stay in the loop"). If the app intends to send promotional messages, the push guide says the soft prompt should say so, because Apple's guidelines require an opt-in before push is used for marketing. It should make declining as easy as agreeing. And it should never pretend to be the system prompt or make the app look broken until the person agrees. A good soft prompt can also point to the settings where a person chooses which kinds of message they get; the tutorial on notification preferences by topic builds that screen and sets its defaults.
Timing and framing are worth testing rather than assuming. The push permission prompt test compares a first-launch prompt with an explanation shown after the first meaningful action, reads opt-in among all new installs (so people who leave before being asked count against the delayed version), and keeps early retention as the guardrail. If too few people are already allowing notifications, the troubleshooting page for a low push opt-in rate separates a prompt problem from a measurement one before anything is rebuilt.
Onboarding messages after the first session
Most people do not reach the activation moment in their first session, and some never open the app again. Onboarding messages exist for them. The push guide sets the order: onboarding messages belong in-app first, with push reserved for people who leave before finishing. One goal per message, the next most useful step, and stop when the person has done it.
In-app messages reach anyone in a session and need no permission, so they carry most of the guidance: a tip on the screen where a feature lives, a card that shows the next step, a reminder of the unfinished booking when the person returns. Push reaches only those who allowed notifications and left, and it should bring them back to the step they did not finish, through a deep link to that screen. Firebase's documentation describes both channels, Cloud Messaging for push and In-App Messaging for in-session campaigns, with event-based triggers, as the push guide states; most engagement tools work the same way, which is what lets messages follow what the person did rather than a calendar.
Where new users also give an email address, the email side follows the same logic. The tutorial on software onboarding emails reaches people when they are not in the product with the one thing that would bring them back, and fits around in-app guidance so the same nudge does not arrive twice. Decide in one place which channel carries each nudge, so a person who stalls does not get the same reminder as an in-app card, a push and an email on the same day. The tutorial on building onboarding messages for new app users turns these rules into a sequence: the steps where each install cohort stalls, one job per message, exits on every step and a holdout.
Measuring onboarding by cohort
Every change to the first session should be read by install cohort: the people who installed in a given week, followed from their first open. Three measures matter together. Activation is the share of the cohort who reached the activation moment within an agreed window; it is a conversion rate (CVRShare of visits or clicks that convert.) through the first-session funnel, and the conversion rate calculator works it out for each step. Permission is the share of the cohort who allowed each permission the change touched, counted against everyone in the cohort, not only those who were asked. Early retention is the share who return at the points the team agreed in advance.
Read them side by side because they trade off. A change that raises opt-in by asking earlier may lose people before they activate; a change that delays the account may raise activation and lower sign-ups that matter later. Log each screen view, each prompt shown and each answer as events, as the deep links guide's event taxonomy recommends, so the drop between any two steps can be seen. Read iOS and Android separately, because the permission models differ. Where possible, run changes as randomised tests rather than before-and-after comparisons, and with small cohorts, treat modest differences as noise until they repeat. For the account step on its own, the deferred account creation test sets out a randomised design that moves only the account and keeps the permission screen where it is.
The customer retention strategy guide treats onboarding as the first retention lever, because early loss is the largest and cheapest opportunity in most businesses. The first-session funnel is where that lever is pulled in an app.
A dog-walking app reorders its first session (illustrative)
Quiltbrack Dog Walks moves its account and permission screens
Quiltbrack Dog Walks, an invented dog-walking booking app, opens with 5 screens before anything useful: a sign-up form, a dog profile, an address form, the notification prompt and a feature tour. Of each weekly install cohort, about 22% book a walk within 7 days, and 38% allow notifications. The team rebuilds the first session so a new user browses walkers near a postcode they type, picks a time and only then creates an account to confirm the booking. The dog profile moves to after the first walk, the tour becomes tips on the screens they describe, and the notification request moves to the booking confirmation, behind a soft prompt that says "we'll tell you when your walker is on the way and when the walk is done. You can choose offers separately in settings." A test over 4 weekly cohorts gives half of new installs the new flow. Bookings within 7 days reach 31% in the new flow against 23% in the old; notifications allowed, counted across every install, reach 44% against 37%, because far more people now reach the moment the prompt appears. Retention at day 30 is 2 points higher in the new flow, within the range the team treats as uncertain at this volume, so it keeps watching it before claiming a gain.
Common mistakes
- Asking for everything at launch. A block of permission prompts before any value spends each single chance at the worst moment. Ask for each permission when its feature is first used.
- An account before any value. Most people can browse and choose without one. Ask for the account when there is something to save, book or buy.
- Soft prompts that hide promotions. If the app will send offers, say so; Apple's guidelines require an opt-in before push is used for marketing.
- A refusal that breaks the app. Push must not be required for the app to function, and no other permission should block the core of the app either. Design the path for people who say no.
- Tours instead of tips. A sequence of feature slides is dismissed unread. Put each tip on the screen where its feature lives.
- Push before in-app. Onboarding guidance belongs inside the session first; push is for people who allowed it and left before finishing.
- Ignoring the deferred link. A person who tapped an ad for one item and lands on a generic welcome has lost the reason they came. Carry the destination through the install.
- Judging opt-in only among people who were asked. Count against every install in the cohort, or a later prompt looks better simply because fewer people reach it.
