Push Notification Frequency: What Works

Web push frequency for ecommerce works best at two to four sends per subscriber per week, with automated triggers counted against the same cap as campaigns. Push has no inbox to sit in and no unsubscribe friction, so an over-sent subscriber does not disengage quietly. They revoke permission, which is far harder to win back.

Count automations and campaigns together

The most common structural mistake is capping campaigns while letting flows send freely on top. The subscriber experiences one stream. If your cart abandonment, back-in-stock and price-drop triggers can all fire in the same day, a subscriber who browses actively can receive six notifications in a week where your calendar says two.

Set one cap per subscriber per week and enforce it across everything. Then decide the priority order for what gets through when the cap is reached, because something has to be dropped. A reasonable order for most catalogues puts transactional and back-in-stock first, cart recovery second, price drop third, and promotional campaigns last, since campaigns are the only category you can reschedule.

Audit this by simulating a single active subscriber rather than reading your calendar. Walk one profile through a browse, an add to cart, a back-in-stock and a price drop in the same week and count what would actually arrive. Most teams are surprised by the total, and the surprise is the point.

Add a minimum gap between any two notifications to the same subscriber. Twelve hours is a workable floor. Without it, a cap of four per week still allows four in one afternoon, which is functionally the same as no cap at all.

What the right number depends on

Catalogue turnover is the main variable. A brand with weekly drops has genuine news four times a month and can support the upper end. A brand with a stable catalogue of twelve products does not, and pushing to fill a cadence produces notifications with nothing in them, which is what teaches subscribers to revoke.

Subscriber recency is the second variable. Someone who opted in this week tolerates and expects more than someone who opted in eight months ago and has not clicked since. Split cadence by engagement recency rather than sending everyone the same volume, and reduce rather than escalate for the unengaged. Escalating on an unengaged subscriber is how you convert a dormant permission into a revoked one.

Device and platform behaviour matters as a practical constraint. Desktop and mobile web push surface differently, and notifications stack differently by browser and operating system. Test what your own subscribers actually see before deciding what a reasonable volume is, because the same cadence reads differently depending on where it lands.

One cadence decision sits outside the cap entirely. Back-in-stock and price-drop notifications are requested by the subscriber for a specific product, so they are permitted even in a week that is already full, and they should be excluded from the cap rather than competing with promotional sends for the same slot. Suppressing a notification the subscriber explicitly asked for is worse for permission than sending one more.

The signals that you are sending too much

Watch permission revocations per send rather than in aggregate, for the same reason the aggregate hides decline as the list grows. A rising revocation rate concentrated on promotional sends is a frequency problem. Concentrated on one trigger is a targeting problem in that trigger.

Watch click rate by position in the week. If the fourth notification of the week consistently underperforms the first three by a wide margin, the fourth is not adding reach, it is spending permission. Cut it and measure revenue for two weeks before deciding.

Watch the gap between delivered and displayed where your provider reports it. Notifications suppressed by the browser or the operating system still count as sends on your side and do nothing for you, and a widening gap often precedes a revocation spike.

Across the programmes we run, [BMO NUMBER: insert the revocation rate difference you measured between capped and uncapped cadence, with the account count and window] is the number we use when a brand wants to add a fifth weekly send. Our e-commerce push notification  [link to: /services/push-notifications] team sets the cap, the priority order and the minimum gap as one rule set, and aligns it with the SMS send caps  [link to: /blog/sms-segmentation] so the same subscriber is not hit twice on two channels in one hour.