App Analytics
Part of Push notification strategy
Measuring opt-outs alongside notification response
Define useful response, track notification preference changes and interpret push metrics with the right denominators and platform limits.
Measure whether notifications help people and whether they respond by turning off a category or changing device settings. Opens alone answer neither question. Pair a defined useful action with message exposure and observable preference changes.
Define response before opening a dashboard
Choose an outcome tied to the message's purpose. A booking-change alert might aim at viewing the revised booking; a reminder at completing the task while it remains useful. Record the eligible audience and observation window before comparing messages. An open can be a step towards the outcome, not the outcome itself.
Separate send attempts, delivery indicators and opens. In Firebase Cloud Messaging Reports, 'Sends' may mean a message was queued or passed to a service such as APNs. 'Received' is an Android-only measure under documented SDK conditions. 'Impressions' covers Android notification messages displayed while the app is in the background.
Reported 'Opens' cover notifications received while the app was in the background. The Reports tab requires Google Analytics, and figures can lag. Check definitions and coverage before naming a ratio a view rate.
Record preference changes accurately
Track these states where the app can observe them:
| State | Record | Use |
|---|---|---|
| Device authorisation and available interactions | Observed state and check time | Shows what the app could use at that check |
| In-app category preference | Category, previous state and new state | Shows which update type the person turned off |
| Message and useful action | Message type, person, exposure definition and outcome window | Connects the intended update with response |
This is a proposed event plan, not a set of events every analytics product supplies. On iOS, people can change notification settings after the first request; on Android, they can change app permission and channel settings. An app can check available settings, but a later check does not reveal the exact moment of an earlier change. Record the check time rather than inventing an opt-out timestamp.
How to Accurately Track Notification Preference Changes
- Observe device authorisation stateRecord current permission status and check time
- Capture in-app category preferenceLog category, previous state, and new state
- Define exposure and outcome windowSpecify message type, user, exposure definition, and observation period
- Avoid timestamp inflationRecord check time—not inferred opt-out time
Match each denominator to its question
A single category's turn-off rate is the number of people in a starting cohort who changed that category from on to off during a period, divided by those in the cohort who had it on at the start.
An exposure-based rate asks how many people who met a clearly defined send or delivery condition later turned it off. Keep those rates separate: a send is not proof that a notification appeared. Deduplicate people, specify the observation window and state which settings changes the app can detect.
A low recorded opt-out count may reflect sparse settings checks or few people enabling notifications. A rise after a product change may reflect an easier-to-find preference control. Compare notification type, exposure, operating system and app version before claiming a cause.
Notification Response vs. Opt-Out Rates: Key Metrics Comparison
- Send
- Message queued or passed to APNs/Firebase Cloud Messaging
- Received (Android only)
- Notification delivered to device under SDK conditions
- Impressions (Android)
- Notification displayed while app is in background
- Open
- User tapped notification while app was in background
- Opt-out (Category Turn-off)
- User changed in-app notification category from 'on' to 'off'
Turn the review into a decision
Read useful action, repeat exposure and turn-offs together. If opens rise without useful action, inspect the message promise and destination; if turn-offs cluster among heavily exposed people, inspect duplicates and frequency; if both action and turn-offs rise, targeting may need review. These are investigation paths, not diagnoses from a single chart.
Record the proposed rule change and compare its results with an appropriately comparable later group.
Key Considerations for Measuring Notification Effectiveness
- Useful Action Rate
- Track actual desired outcomes (e.g., booking view, task completion)
- Exposure-Based Opt-Out Rate
- People who turned off a category after meeting exposure criteria
- Frequency Sensitivity
- High exposure may correlate with increased opt-outs; review targeting
- Platform Differences
- iOS allows post-request setting changes; Android has channel-level controls
