Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Mobile & App Marketing › Troubleshooting

Users Turning Off App Notifications

Why app users switch notifications off and how to fix it: no shared cap, broadcasts to everyone, one switch, messages at night and taps that go nowhere.
Troubleshootingintermediate10 min readUpdated

First, locate the fall

Start from three signals rather than one. The push notifications and in-app messaging guide names notification disables, topic opt-outs and uninstalls in the days after a send, together with a falling enabled rate, as the leading indicators that the channel is being overused. Read the enabled rate by install cohort so that a change in the mix of new and long-standing users does not move it on its own, and keep iOS and Android apart, because their permission histories differ.

Then attribute every disable, opt-out and uninstall to the last notification the person received before it. Count them four ways: by message, by the team or tool that sent it, by topic in the app's settings, and by the local hour it arrived. A cluster after one campaign points at that campaign's audience or wording; a cluster after one team's sends points at the missing shared cap; a cluster in a single topic points at what that topic now carries; and a cluster in the small hours points at timing. If disables rise steadily across every message with no peak, suspect volume and the absence of topics before any single push.

Two boundaries keep the diagnosis honest. A first request that few people accept is a different problem, handled by the diagnosis of a low push notification opt-in rate; this page assumes people did allow notifications and are now taking that permission back. And people who stopped opening the app altogether are a lapse problem for the guide to re-engaging lapsed app users, not a notification fault.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Disables climb in weeks when several teams run campaigns at onceNotifications received per person per period, summed across every tool that sendsNo frequency cap shared across every team and campaign that can sendOne cap per person across all senders, with a priority order and transactional messages outside it
Opt-outs and uninstalls peak the day after scheduled sends to the whole baseThe share of sends caused by a person's action against sends on a calendarBroadcasts sent to everyone instead of messages triggered by behaviourMake triggered messages the majority and segment what remains
People switch off notifications entirely, then contact support about missed order updatesWhether the app offers topics, and which message preceded each full disableOne switch for everything, so annoyance with offers turns off service messages tooTopic preferences, with service topics kept apart from offers
Disables concentrate on messages delivered late in the evening or overnightDelivery hour in each recipient's own time zone for messages followed by a disableNotifications arriving at night in the recipient's own time zoneQuiet hours as a system rule, with non-urgent sends held until morning
Taps are recorded but no session reaches the promised itemDeep link opens against the screen views and item loads that followTaps that land on the home screen or a broken route instead of the promised screenRoute every link to its exact screen, with fallbacks for signed-out people and missing items
A curious line draws taps, then disables follow within the same sessionWording of each message against what the opened screen actually showsVague or teasing copy that the next screen does not deliverLead with the specific change and promise only what the tap will show

Fix, verify and prevent each cause

Every team sending on its own budget

Product nudges, marketing offers, service alerts and content updates are often sent by different people through different tools, and each looks modest when judged alone. The push guide sets the remedy: a Frequency cap per person, per period, across every campaign and every team that can send, with a priority order so that when several messages qualify the most important goes and the rest wait or drop.

Fix: inventory every sender, agree one cap and one priority order, keep transactional messages outside the cap and put promotional ones at the bottom.

Verify: the number of notifications any single person receives in a period stops exceeding the agreed cap in every tool's logs, and disables in the busiest campaign weeks fall back towards the quiet-week level.

Prevent: make the cap a setting of the sending platform rather than a line in a campaign brief, and require any new tool that can send to read it before it goes live.

A calendar that sends the same message to everyone

A notification caused by something the person did, or by something that changed for them, is relevant by construction. A notification sent because a date arrived is relevant only by luck. The push guide reports that triggered messages almost always outperform broadcasts on both response and tolerance, and asks that broadcasts be the minority, segmented and counted against the same cap.

Fix: list the scheduled sends, replace each one that has a behavioural equivalent with a trigger (a saved item back in stock, a reply, a booking soon due), and narrow the remainder to the segment the news actually concerns.

Verify: run a holdout of eligible users for each replaced message and compare the target action and the disables between groups; the triggered version should hold or raise the action while costing fewer opt-outs.

Prevent: require every new scheduled send to name its segment and to explain in one sentence why a trigger cannot do the job.

No way to say "less of this, more of that"

When the only control is the device's own switch, a person irritated by offers has one way out, and it silences the delivery update they were waiting for as well. The tutorial on notification preferences by topic builds the alternative step by step, and notes that a rising opt-out on one topic is a specific, fixable signal while a falling enabled rate is a late and vague one.

Fix: follow the tutorial's inventory, classify every message, group them into topics people recognise, start promotional topics off unless chosen, and make every sending tool read the same preferences before each send.

Verify: after launch, topic opt-outs should rise while full device-level disables fall, showing that people now adjust rather than abandon; test by switching a topic off on a test account and triggering every tool.

Prevent: give each topic an owner who approves new messages into it, so offers never creep into the topic everyone keeps on.

Messages that wake people up

A well-written notification at the wrong hour still costs more goodwill than it earns. The push guide is explicit that quiet hours should be a system rule rather than a campaign setting, that scheduling should use each recipient's local time, and that transactional messages can pass because the person expects them. Campaigns scheduled in the head office's time zone are the usual culprit when an app's audience spreads across regions.

Fix: store each user's time zone, schedule in local time, and hold non-urgent sends that fall inside quiet hours until the morning.

Verify: the share of promotional notifications delivered inside quiet hours drops to nothing in the delivery log, and the overnight cluster of disables disappears from the next cohort's figures.

Prevent: add a delivery-hour check to the pre-send review, and alert the messaging owner whenever a non-transactional message is queued for delivery inside quiet hours.

Taps that open the wrong screen

A person who taps a notification about their order and lands on the home screen has learned that tapping is pointless, and some will decide that the notifications are too. Every notification should carry a Deep Link to the exact screen with the item loaded, and the deep links and app analytics guide explains the route table that maps links to screens and the awkward cases it must handle: a signed-out person, an item that no longer exists, a screen renamed in a release.

Fix: route each message type to its screen, return people to the intended screen after sign-in, and design a sensible alternative when the item is gone.

Verify: compare deep link opens with the screen views that follow them, per message; the gap between a tap and the promised screen should close, and the tutorial on testing app deep links gives the matrix of app states and versions to run before the next send.

Prevent: retest every live notification route against each release, because a renamed screen breaks links silently.

Words the app does not honour

"Something special is waiting" earns a tap once. When the opened screen holds nothing special, the person reads the next notification with suspicion or switches the channel off. The tutorial on writing push notifications asks for the change to lead the title, the specific thing to be named, the tap's result to be stated, offers to be labelled as offers, and the words to be judged on the action after the tap rather than on opens.

Fix: rewrite the offending messages around what changed for this person and what the tap will show, and remove any line that relies on curiosity alone.

Verify: test the new wording against the old inside a holdout, and count disables and topic opt-outs in the following days as costs beside the target action.

Prevent: keep the tutorial's checklist in the send approval, so nobody can schedule a message whose promise has not been checked against its destination screen.

A prevention routine for the message programme

Treat the permission people already gave as a balance that every send draws on. The push opt-in rate measures how much of it the app earned at the first request; the enabled rate among existing users measures how much is left. Put the second figure, by platform and install cohort, on the same weekly report as the target actions the messages exist to drive, and have one owner for the whole programme rather than one per team.

Each week, read disables, topic opt-outs and uninstalls joined to the preceding message, ranked by message and by sender. Each month, compare scheduled sends with triggered ones, and walk the delivery log for anything sent inside quiet hours. Before any new message type launches, place it in a topic, confirm it counts against the shared cap if it is promotional, test its route on current app versions, and have its words checked against the screen it opens. For recurring messages, keep a standing holdout, so the programme can always show which sends add activity and which only spend permission.

Apple's App Store Review Guidelines say, as the push guide records, that push must not be used for marketing unless the person has opted in, which makes the topic structure a compliance safeguard as well as a retention one. Where some of these messages could also go by text, the comparison of SMS and push notifications sets out which channel each belongs on, so the same reminder does not arrive twice. Other app symptoms, from costly installs to a store page that converts less, are listed in the app marketing troubleshooting section.

Illustrative example — not a real result

Gazzlenook Events finds its weekend problem

Gazzlenook Events, an invented listings app for local gigs, markets and festivals, sees the share of active users with notifications on slide from 61% to 48% over a quarter. Joined to the preceding send, 70% of disables follow Friday or Saturday messages. Two teams send on those days: editorial pushes a "this weekend" roundup to everyone, and partnerships pushes sponsored ticket offers, with no cap between them, so heavy users receive up to five notifications in two days. A fifth of disables follow messages delivered after 23:00 local time, because both teams schedule in the head office's zone. The app has one switch. Gazzlenook sets a cap of three promotional notifications a week across both teams, replaces the roundup with alerts for venues and artists a person follows, launches topics (your tickets, things you follow, weekend picks, offers) with offers off unless chosen, and holds non-urgent sends between 21:00 and 09:00 local time. Over the next six weekly cohorts, weekly disables fall from 2.1% to 0.6% of enabled users, 18% of people switch offers off while keeping ticket updates, and the enabled rate recovers to 55%.

Common mistakes

Rewording a message when the real fault is that five teams send it to everyone. Reading the enabled rate across the whole base, where years of older users hide the latest cohorts. Treating a cut in volume as proven without a holdout. Adding topics that one sending tool ignores. Asking people to re-enable notifications before fixing what made them leave.

Frequently asked questions

How do we know which messages are making people switch notifications off?

Join each disable, topic opt-out and uninstall to the last notification the person received before it, and count them by message, sending team, topic and local hour of delivery. The push guide lists those three signals, together with a falling enabled rate, as the leading indicators of overuse. A spike that follows one campaign or one team's sends is a specific finding; a slow decline across every message usually points at the volume or the missing topics rather than any single push.

Should we send fewer notifications to win people back?

Fewer is usually part of it, but the order matters. First put every sending team under one cap with transactional messages outside it, then replace broadcasts with triggered messages, then hold non-urgent sends to daytime in each person's local time. Prove each change with a holdout, as the push guide describes, because cutting volume blindly can also cut the messages that people wanted and that brought them back.

Can we ask people to turn notifications back on?

Only from inside the app. On iOS the system prompt cannot be shown again once it has been answered, so a person who switched notifications off can only restore them in the device settings. The preferences tutorial suggests saying so plainly on the app's settings screen and explaining how to turn them back on, alongside topic switches that let the person keep order updates while leaving offers off. Make sure the app still works well without notifications, as Apple's guidelines require.

Terms used on this page

Where this page is used

Also related

Sources

  1. App Store Review Guidelines — Apple Developer

Related