Push notifications are the only marketing channel where the customer can revoke your access permanently with two taps, and where doing so takes less effort than unsubscribing from an email. That asymmetry is why push notification best practices look so different from everything else in the stack, and why most brands quietly waste the channel.
This guide covers how push differs from email and SMS, how to earn the opt-in, what a good notification actually looks like, how to manage frequency before people mute you, and which metrics tell you whether the program is healthy.
Why Push Is Different
Push is high-visibility, low-tolerance, and irreversible when it goes wrong. It appears on a lock screen with no subject line to skim past and no folder to file it in. There is no promotions tab to absorb a mediocre message, and no equivalent of a re-permission campaign once someone disables notifications at the OS level.
The practical consequence is that push should carry your most timely and most personally relevant messages, and nothing else. Every notification you send that could have waited for an email costs you a little of the permission you will need later.
Push, Email, SMS, and In-App Compared
| Factor | Push | SMS | In-app message | |
|---|---|---|---|---|
| Reach | Opted-in app users only | Broadest | Opted-in, high delivery | Active session only |
| Immediacy | Very high | Low to medium | Very high | Immediate, contextual |
| Cost per message | Effectively zero | Very low | Meaningful per message | |
| Tolerance for volume | Low | Medium | Very low | High, user is already there |
| Best for | Time-sensitive, personal triggers | Depth, education, promotion | Urgent, transactional | Onboarding, feature discovery |
| Recovery if opted out | Very difficult | Straightforward | Requires new consent | Not applicable |
Reading that table sideways is the whole strategy. Push and SMS are for moments. Email carries the depth. In-app handles anything that only makes sense while someone is using the product. Coordinating the four is what omnichannel marketing is actually for.
Getting the Opt-In Right
Never ask on first launch
The default implementation triggers the system permission prompt the moment the app opens, before the user has any reason to trust you. This is the single most common and most expensive mistake in mobile marketing, because a denial at the OS level is close to permanent.
Use a soft ask first
Show your own in-app screen explaining what notifications you send and why they are useful. Only users who tap yes on your screen see the system prompt. Users who decline your screen can be asked again later. This pattern routinely doubles opt-in rates because you spend the one-shot system prompt only on people already inclined to accept.
Ask at a moment of demonstrated value
After a first purchase, after saving a favorite, after completing setup, or when the user takes an action where a notification is obviously the payoff. “Want us to tell you when this ships?” converts far better than a generic request for permission.
Be specific about what they will get
Name the notification types and let people choose categories where you can. Granular preferences reduce total opt-ins slightly and increase retained opt-ins substantially, which is the trade you want. It is the same consent logic that governs email and SMS permission.
What Makes a Good Push Notification
Copy that works in fifty characters
Assume the title is truncated around forty characters and the body around one hundred, less on a lock screen. Front-load the specific thing that happened or the specific benefit. Lead with the noun, not the greeting. “Your order shipped” beats “Hi Sarah, we have an update for you.”
Personal beats promotional
The notifications that keep opt-ins alive reference something the user did or owns: an item they saved, a subscription running low, a reward about to expire, a reply to something they posted. Broadcast promotions are the fastest route to a disabled channel, so ration them carefully.
Deep link to the exact screen
If a notification about an item opens the home screen, you have wasted the tap and taught the user that your notifications are not worth acting on. Every push should land on the specific content it references, and this is worth auditing quarterly because deep links break silently during app updates.
Time it to the person, not the calendar
Send in the recipient’s local time, respect quiet hours, and use per-user send-time optimization where your platform supports it. A notification at two in the morning does more damage than a week of silence.
Frequency and Fatigue
There is no universal correct frequency, but there is a reliable diagnostic: watch opt-out rate and open rate together, segmented by how many notifications the user received in the previous week. The point where opens start falling faster than volume rises is your ceiling, and it is usually lower than marketing wants and higher than engineering fears.
Set global frequency caps that no individual campaign can override, add a minimum gap between notifications, and give promotional messages a lower priority than triggered ones. Then hold the cap even during peak season, because the revenue from an extra December send is smaller than the lifetime cost of the opt-outs it produces.
The Types of Push That Earn Their Keep
- Transactional and status updates. Orders, shipping, deliveries, appointments, and account activity. Highest engagement, lowest complaint rate.
- Abandonment recovery. Cart and browse reminders, sent once, with a genuine reason to return rather than a discount reflex.
- Replenishment and renewal. Timed to consumption or subscription cycles, which is where predictive timing pays off directly.
- Back in stock and price drop alerts. Explicitly requested, so treat them as promises rather than campaigns.
- Loyalty and reward events. Points earned, tier reached, reward expiring. Genuinely welcome because the user has a stake.
- Onboarding nudges. A small, finite series that gets a new user to their first meaningful action, then stops.
What is conspicuously absent from that list is the daily promotional blast. If your push calendar looks like your email calendar, the channel is being misused.
Segmentation and Triggers
Push works best as an output of behavior rather than an item on a schedule. That means your segmentation needs to run on events, not lists: last session date, items viewed, purchase recency, feature adoption, and lifecycle stage. The principles carry over directly from segmentation and personalization, with the added constraint that push punishes irrelevance far more harshly.
One practical rule from running these programs: build the trigger library before the campaign calendar. Teams that start with triggered notifications and add broadcasts later maintain healthy opt-in rates. Teams that start with broadcasts spend the following year trying to recover them.
Measuring Push Performance
Track opt-in rate by acquisition source, delivery rate, direct open rate, influenced conversions within an attribution window you agree on in advance, opt-out and notification-disabled rate, and uninstall rate correlated with notification volume. That last pairing is the one most teams never look at and the one that reveals whether the channel is quietly costing you users.
Push also deserves incremental testing rather than raw attribution. Hold back a portion of the eligible audience and compare, because a great deal of push-attributed revenue would have arrived anyway. The same measurement discipline covered in data-driven decision making applies with extra force here.
Common Mistakes
- Requesting permission on first launch, before any value has been demonstrated.
- Sending the same message to every user because the platform makes it easy.
- Deep links that dump users on the home screen.
- Ignoring time zones on a global audience.
- Treating push as a free channel because there is no cost per message.
- No frequency cap, so several campaigns stack on the same user in one afternoon.
- Never measuring uninstalls against notification volume.
Frequently Asked Questions About Push Notifications
What is a good push notification opt-in rate?
It varies by platform and category, but a well-implemented soft-ask flow generally lands well above what a first-launch system prompt achieves. Track your own baseline by acquisition source and improve against it rather than chasing a benchmark.
How many push notifications per week is too many?
Find it empirically. Segment opens and opt-outs by notifications received in the prior week and look for the point where opens fall faster than volume rises. For most consumer apps the sustainable ceiling is a handful, not a daily send.
Should push and email carry the same message?
Rarely. Push should carry the moment and email the detail. Sending both for the same broadcast promotion doubles the annoyance without doubling the revenue.
Can users be won back after disabling notifications?
Only through in-app messaging or another channel, and only by asking them to change an OS setting, which very few will do. Prevention is the entire strategy.
Do I need a mobile app for push at all?
Web push exists and works for some categories, but engagement is typically lower and browser support is inconsistent. The case for owning an app is a separate one, covered in the importance of having a mobile app.
Final Thoughts
The best push notification best practices all reduce to one principle: treat the permission as borrowed and expensive. Earn the opt-in at a moment of real value, send messages that reference something the person actually did, cap frequency and hold the cap, deep link every tap, and measure uninstalls alongside conversions. Restraint is what keeps this channel working, and restraint is exactly what the tooling makes it easy to abandon.