Push Notifications Have a Spam Problem That Developers Built
Push notifications were introduced as a mechanism for delivering timely, relevant information to mobile users. They have become, in the hands of most apps, a mechanism for re-engaging users who have stopped using an app — delivered at volumes and frequencies that have trained users to disable notifications as a reflex rather than a deliberate choice.
The data on notification effectiveness tells a story that most growth teams choose not to hear. Opt-in rates for push notifications have declined steadily as users have learned from experience what push permission grants apps permission to do. iOS’s explicit permission prompt — which apps must request before sending any notification — shows opt-in rates below 50 percent for most app categories. Users who do opt in disable notifications at rates that correlate directly with how many notifications an app sends, not with how relevant those notifications are.
The Re-Engagement Trap
The product team logic that drives notification abuse is coherent from a certain angle: users who are not opening the app are not being monetized; a notification that brings them back produces immediate, measurable value; the cost of the notification is zero. This logic ignores the deferred cost — the user who disables notifications, the user who uninstalls the app, the user who leaves a negative review mentioning notifications specifically — which is real but not attributed to the notification decision in most analytics setups.
The attribution failure is not accidental. Session-based analytics measure what brings users in. They do not measure what pushes users out. A user who disables notifications and then stops opening the app appears in the analytics as a user with declining engagement — the notification strategy appears to have failed to retain them, when in reality the notification strategy caused the retention failure. The measurement system produces an interpretation that exonerates the strategy that created the problem.
What Works
Notifications that users rate as valuable share identifiable characteristics. They are timely — they arrive when the information is relevant, not on a promotional schedule. They are specific — they tell the user something they did not already know or would have had to open the app to discover. They are actionable — the user can respond or use the information without necessarily opening the app at all. And they are infrequent enough that each notification carries signal value rather than becoming ambient noise.
Transactional notifications — order shipped, payment received, reservation confirmed — meet all of these criteria and achieve opt-in rates and interaction rates that promotional notifications cannot approach. Users want to know when something they care about has changed. They do not want to be reminded that the app exists or that a sale is happening.
The notification personalization that machine learning enables — sending notifications based on individual user behavior patterns rather than on product manager schedules — improves outcomes relative to broadcast notifications. An app that has learned when a particular user is likely to engage and sends notifications during those windows performs better than an app that sends to all users on a fixed schedule. The improvement is real but does not overcome the fundamental problem that most notifications are sent not because the content is valuable to the user but because sending it is convenient for the business.
The iOS 15 Effect
Apple’s notification summary feature — which groups non-time-sensitive notifications into a scheduled delivery rather than delivering them immediately — revealed something important about how users experience notification volume. The users who adopted notification summary reported higher satisfaction with the apps they used, not lower. Batching notifications that are individually low-priority into a single daily or twice-daily summary reduced the interruption cost without reducing the information value.
This is the most honest signal the industry has received about optimal notification frequency: most of what apps send is not time-sensitive, most users prefer receiving it less frequently, and the app experience improves when the interruption pressure decreases. The apps that treat this as an invitation to send more frequently to overcome the summary mechanism are making a calculation about short-term re-engagement that destroys long-term notification permission.
Send notifications when you have something to say. The alternative is teaching users that your notifications have nothing to say.