One limit cannot describe every message
An order-status alert and a general promotion create different expectations. Put messages into documented categories before setting a cap.
Operational
Security events, account changes, delivery status, and requested reminders. These should be event-driven and deduplicated, not used as a route around promotional limits.
Editorial or product updates
New content, saved-search results, product availability, or release notes. Frequency should follow the promise made when the person subscribed.
Promotional
Offers and broad campaigns consume attention fastest. A practical starting default is no more than one promotional notification per subscriber per day, with stricter weekly limits for inactive audiences.
The suggested starting limits in this guide are operating recommendations, not official numerical limits from Google, Apple, Mozilla, or the Web Push standard.
Calculate eligibility at send time
A campaign audience should be reduced by permission state, explicit opt-outs, message-category preferences, recent sends, quiet hours, invalid endpoints, and account suppression rules. The push notification segmentation guide shows how to document those inclusion, exclusion, and freshness rules.
Use a rolling window
A calendar-day cap can send at 11:55 p.m. and again just after midnight. A rolling 24-hour window better reflects what the subscriber experiences. Store timestamps in UTC and apply local quiet hours separately.
Reduce pressure as engagement fades
- Prioritize active subscribers who recently clicked or returned.
- Lower promotional frequency after 30 to 60 days without engagement.
- Pause promotional sends after a longer inactive period.
- Keep requested operational events independent, but deduplicate them.
- Provide category preferences before forcing a full unsubscribe.
Protect campaigns from bursts
The service's current delivery model applies a rolling repeat-contact boundary and hard account allowances. Treat those as product safeguards, then define stricter message-category caps when the subscriber promise requires them.
Frequency caps protect the subscriber, while rate limits protect the system. Queue bulk deliveries, spread work across time, and use the Web Push status-code and retry decisions to distinguish throttling from permanent failures. Idempotency keys should prevent a retry from becoming a duplicate notification.
Watch the metrics that reveal fatigue
Track opt-outs, browser permission loss, notification closes when available, clicks per accepted message, return visits, and downstream outcomes. Compare these by recent-send count and subscriber activity using consistent push notification metric definitions. If volume rises while useful outcomes stay flat and opt-outs increase, the extra messages are not helping.
A policy a team can operate
- Define categories and who owns each one.
- Set rolling caps by category and combined promotional volume.
- Document quiet hours and time-zone fallbacks.
- Calculate eligibility immediately before queueing.
- Record every suppression reason.
- Review inactive-audience rules monthly.
Resolve conflicts with an explicit priority order
A subscriber can qualify for several messages at once. Rank categories before this happens. Security and requested account events may take priority, followed by time-sensitive saved-item alerts, editorial updates, and promotions. When two messages have the same priority, prefer the newer event or combine them only when the combined notification remains specific and actionable.
Record the message that won and the messages suppressed. Otherwise a team may interpret lower send volume as a broken campaign instead of a working attention policy.
Use a worked rolling-window check
Assume a subscriber received one promotion 18 hours ago and the policy allows one promotional notification in any rolling 24-hour period. A new promotion is suppressed for six more hours. A requested password-change alert can still send because it belongs to a separately defined operational category. These numbers illustrate the calculation; each product should choose and review its own limits.
Use ethical guidance as the minimum boundary
MDN’s Web Push best-practices guide describes notifications as appropriate when they are useful and time-sensitive and emphasizes easy opt-out. A frequency policy turns that principle into an auditable send-time decision. Place that policy inside the broader push messaging and growth framework so writing, segmentation, limits, and measurement remain connected.