The average smartphone user receives 46 push notifications every day. Most go unread. A meaningful number causes the user to open their settings and silence the app that sent them. If you are running a mobile app, there is a real chance your notifications are working against you. Not because the channel is broken, but because most push strategies are built around volume rather than relevance. This guide covers what actually drives retention through push, how to segment users properly, how to read the data that tells you whether it is working, and where deep linking fits in.
How push became a broadcast channel (and why that is a problem)
Push notifications were designed as a utility channel. Their job was to deliver something time-sensitive and genuinely useful: your order has shipped, someone replied to your message, your flight is delayed.
Over time, they became a marketing tool. Teams started sending weekly promotions, re-engagement blasts, and "we miss you" messages to every lapsed user regardless of why they had gone quiet. The logic was simple: send more, get more opens.
The actual result: users who feel a channel is noisy revoke permission. And once push permission is gone, it almost never comes back.
Notification permission is not a marketing asset. It is a trust signal. The moment you treat it like a broadcast channel, users start treating you like noise.
What the data says about push performance
The channel works. Just not the way most teams use it.
The gap between triggered and broadcast notifications is significant. A behaviour-based push converts at three to six times the rate of a scheduled campaign sent to your full user base. The reason is simple: triggered notifications are relevant to what the user is actually doing. Broadcasts are relevant to what the marketing team decided to send that week.
Step one: earn the opt-in before you ask for it
iOS requires explicit permission before you can send push notifications. Most apps present the permission prompt immediately on first launch, before the user has any reason to say yes.
High-performing apps delay the prompt until after a moment of genuine value. A food delivery app that asks for push permission after a user's first order, framed around order status updates, converts at a higher rate than one that asks on the welcome screen. A fitness app that waits until a user completes their first workout and frames notifications around workout reminders sees the same effect.
The test is simple: can the user answer the question "What will I actually get from these?" before you ask? If not, you are prompting too early.
Segmentation: going beyond the basics
Sending the same notification to every user is the single biggest driver of opt-out. Your user base is not a single cohort, and treating it as such produces results that reflect that.
Most teams start with lifecycle segmentation: new, active, lapsing, churned. That is the right foundation. But the teams that see the biggest lift in push performance go further.
Behavioural segmentation
Group users by what they actually do in your app, not just how recently they opened it. The user who has opened your app 40 times but never completed a purchase is a different person from the one who purchased twice and went quiet. They need different messages.
Feature adoption: users who have used a feature once vs those who use it regularly respond to entirely different prompts.
Session depth: users who open the app and leave in under 30 seconds are not getting value. A notification about a different entry point may change that.
Incomplete actions: started onboarding but did not finish, added to cart but did not buy, began a form but did not submit. Each is a distinct segment with a clear trigger point.
Preference-based segmentation
Give users control over what they receive and use their choices as data. An app that lets users select notification categories (deals only, account updates, new content) learns what each person actually wants. That preference data becomes a segmentation signal.
Users who opt into specific categories also tend to engage with those notifications at significantly higher rates than users who receive everything by default. Giving users control reduces opt-outs and improves the quality of your remaining audience.
Engagement history segmentation
Track how individual users respond to push over time. A user who has tapped 8 of your last 10 notifications is in a different state than one who has not tapped any in 60 days. Treating them the same wastes your highest-value audience and burns the patience of your lowest-value one.
Highly engaged: can receive more frequent, richer notifications. Test new formats here first.
Occasionally engaged: stick to high-signal triggers only. Reduce frequency.
Non-engaged: try one well-targeted re-engagement notification before suppressing. If they do not respond, remove them from regular sends to protect your opt-out rate.
Predictive segmentation
If your data volume supports it, build segments around predicted behaviour rather than observed behaviour. Users who match the behavioural pattern of past churners are a segment worth targeting with retention-focused messages before they actually churn. Users who match patterns of high-LTV customers are worth treating differently in onboarding.
This does not require a machine learning team. Most modern push platforms expose flags for at-risk and high-value users based on standard behavioural models. The value is in using those segments rather than sending to everyone.
Segmentation is not about sending fewer notifications. It is about sending the right notification to the right person at the right moment. Volume follows from that; it does not lead it.
Trigger on behaviour, not the calendar
The most effective push notifications are not scheduled. They fire because something happened, or stopped happening.
User completed onboarding but has not used the core feature after 48 hours
User added items to a cart and did not complete the purchase within 24 hours
User has not returned for 5 days after previously opening daily
User reached a milestone and has not come back the next day
A time-sensitive event relevant to that specific user becomes available
Each of these arrives at a moment when it is genuinely relevant. None of them requires a campaign brief or a send schedule. They fire based on user behaviour and stop when the user takes the intended action.
If triggered sends are under 30 percent of your total notification volume, that ratio is worth examining. It usually means your push strategy is still built around campaigns rather than context.
The detail most teams miss: deep linking from notifications
A push notification that taps through to your app home screen is a missed conversion. If the notification says "Your discount expires tonight," the tap should open the cart, not the home screen.
Deep linking from notifications connects the message to the exact in-app destination in one step. No extra navigation. No searching. Just the action the user needs to take.
For re-engagement campaigns, getting this right moves conversion rates by 20 to 30 percent. Every navigation step between notification tap and target action is a drop-off point.
Metrics: how to read the data and actually act on it
Open rate is what most dashboards show first. It is also one of the least actionable metrics in push. A 15 percent open rate that leads to no meaningful in-app action has produced nothing. A 6 percent open rate that drives 80 percent of openers to complete a purchase is performing well.
Here are the four metrics that tell you what is actually happening, and what to do when each one moves.
How to close the feedback loop
Metrics are only useful if they change what you do next. A practical rhythm for most teams is a weekly review of opt-out rate by send type and conversion rate by segment, and a monthly review of retention and revenue trends by push engagement cohort.
When the opt-out rate spikes after a specific send, trace it back: what was the segment, what was the frequency that week, what did the message say? Usually, the culprit is one of three things: too many sends in a short window, a message that felt off-topic for the segment, or a notification that arrived at the wrong time of day.
When the conversion rate drops on a trigger that previously performed well, the usual causes are a copy change that weakened relevance, a broken deep link, or a segment that has drifted (the users you are targeting have changed, but the trigger definition has not).
The goal is not to optimise each metric in isolation. The goal is a push channel where opt-outs stay flat, conversion stays healthy, and the audience that trusts your notifications grows over time.
A five-point audit for your current setup
Opt-out rate trend. Is it rising? Which send types correlate with spikes?
Segmentation in your last five campaigns. Were they sent to the full user base or defined segments?
Deep link coverage. Where does a notification tap actually land for users?
Sends per user per week. Pull the distribution. Are some users receiving 15 or more notifications in a week?
Triggered vs scheduled split. If triggered sends are under 30 percent of total volume, the balance is worth reviewing.
Push works when it earns its place on a user's lock screen. That means relevant segmentation, behaviour-based triggers, clean deep linking, and metrics that feed back into the next decision. Every piece depends on the others. Teams that treat them separately end up optimising the wrong thing.
Linkzly has a free plan if you want to see how push notifications, deep linking, and delivery analytics work in practice. No credit card required, set up in under 30 minutes. linkzly.com