Why publishers still run push ads after browsers restricted notifications

Push ads deliver notification-style creatives straight to a lock screen or desktop tray, skipping the banner slot. A visitor accepts a permission prompt once, and a service worker keeps sending messages after leaving the page, which is why the format outlasted years of ad blocker growth while display CPMs collapsed. Chrome and Firefox tightened the subscription flow in 2020, cutting weak bases and raising costs, but volume held in gambling, dating and finance offers. The mechanics, the pricing, the formats that replaced classic push, and the mistakes that drain a budget by day three follow.

What push ads actually are and how the opt-in works

A push notification ad is not slotted into a page the way a banner or a native widget is. It rides the same channel a site uses to tell a returning visitor about a restocked item: the Push API built into Chromium and Firefox, paired with a service worker that keeps running after the tab closes. The subscriber grants permission once, through a browser-level prompt the publisher does not design, and every message after that lands directly in the operating system's own notification tray, which is the entire reason media buyers still call the format push ads.

The clearest parallel I have found for that permission fatigue sits, oddly, on Repères Collège, a page that tracks how French middle-school reforms get announced and then quietly shelved: a proposal produces the same short behavioural spike a notification prompt does, and both fade once the routine sets in.

Subscription rates behave the same way in a media-buying context. A fresh zone converts at one rate on day one and at a fraction of that by week three, regardless of which creative is running, simply because the novelty of the prompt itself wears off for a repeat visitor.

Nothing about the format requires the advertiser to own inventory. A buyer picks a zone, sets a bid, and the exchange decides which subscribed device sees the notification next, based on the bid and the remaining daily cap on that zone. That auction layer is what turned push from a retention tool built for e-commerce into a traded ad format with its own networks and self-serve dashboards. Retail sites used the channel to bring lapsed shoppers back; ad-tech turned the same mechanism into inventory anyone could bid on.

How push ads are targeted, priced and bought

Targeting sits closer to search than to display: GEO, OS, device type, carrier, browser version and connection speed are the usual filters, stacked before creative rotation and frequency capping decide what a given subscriber actually sees. A campaign aimed at a Tier 1 GEO with an unrestricted OS filter draws a different price than one narrowed to a single carrier in a Tier 3 market, and the gap between the two can run tenfold on identical push ads inventory.

Buyers who skip the carrier filter on mobile-heavy GEOs usually discover why it exists after the first invoice, once spend quietly concentrates on a handful of low-quality connections that convert clicks but not signups.

Pricing runs on CPC in most self-serve panels, with CPM reserved for branding runs and CPA offered only by networks willing to absorb more fraud risk on the advertiser's behalf. I first saw the CPC-versus-CPM split explained clearly on push notification ads, in a breakdown of why volume buyers avoid CPM once a base ages past its first month. Bid steps of a fraction of a cent look trivial until multiplied across a million daily impressions on a broad zone, and that arithmetic is why panel dashboards push new advertisers toward CPC by default rather than push ads running on flat CPM.

Which verticals still get real return from push ads

Display CPMs collapsed across most consumer categories once ad blockers reached majority adoption on mobile, but a handful of verticals kept buying notification inventory at stable or rising bids, which is exactly the behaviour push ads were built to interrupt.

What they share is a short decision cycle: the offer converts within minutes of the click or it does not convert at all, which suits a format that interrupts rather than persuades a visitor across several sessions. Two groups account for most of that spend, and each buys for a different reason. I traced the split by comparing zone-level reporting across three networks over one quarter, sorting placements by vertical instead of by GEO.

Dating and casino offers

Dating and real-money gaming offers convert on urgency and novelty, both of which a notification delivers better than a banner sitting in a sidebar the visitor has already learned to ignore. Casino operators buying this inventory typically rotate creative on a 48-hour cycle, since a static message loses roughly half its click-through rate after the second day on the same subscriber base.

Utility and finance offers

Loan and antivirus offers work through a different mechanism: the notification mimics a system alert closely enough to earn a click from a subscriber who is not actively shopping. Networks that police creative quality reject the more aggressive versions of this pattern, which pushes that spend toward panels with looser review, and explains why finance-vertical CPCs vary more between networks than casino or dating CPCs do.

Where push ads formats diverge: native, in-page and calendar

The 2020 to 2022 permission crackdown split what used to be one format into three. Native push still exists, still needs the OS-level opt-in, and still cannot reach iOS Safari users at all, since Apple did not ship Web Push support until late 2022, and adoption of it inside ad-tech buying stacks lagged well behind that release. In-page push filled the gap with a creative styled to look like a system notification but rendered inside the page itself, requiring no permission at all and therefore reaching every device the site itself reaches, which is the version of push ads most GEO-broad campaigns run on today.

The distinction matters at the buying stage, and it is worth reading through before allocating budget to any single push notification ads placement, since the two formats convert on different mechanisms and rarely optimise the same way inside one campaign.

Calendar push, built on malformed ICS invites that auto-subscribed a device to a recurring reminder, is close to extinct after Apple and Google patched the calendar-app behaviour it depended on. A handful of panels still list it at a steep discount, but delivery on current iOS and Android builds is close to zero, and that discount usually means dead inventory.

What breaks a push ads campaign before it earns anything

Most underperforming runs fail for one of two structural reasons rather than a weak creative, and both show up in the reporting within the first few hundred dollars of spend if a buyer knows where to look, long before a full push ads budget is committed.

The first is buying a stale subscriber base without checking its age distribution, and the second is skipping invalid-traffic filtering on a network that does not enforce it by default. Reading the zone documentation on push-ads.io before committing budget catches a surprising share of both problems, since reputable networks disclose base age bands and IVT filtering status directly in the placement panel rather than burying it inside a follow-up support ticket nobody reads.

Subscriber base decay

A push subscriber list is not a static audience. Devices get reset, browsers get reinstalled, and users unsubscribe once a run gets too aggressive with frequency, so a zone that converted well six months ago is not the same zone today even if the reported size has not changed. Buyers who request a base refreshed within the last 30 to 60 days consistently see better click-through than those who buy on raw size alone.

Invalid traffic and bot farms

Subscriber farms that auto-accept the permission prompt on emulated devices exist specifically to inflate zone size, and they cluster in the same low-cost GEOs that also carry the cheapest legitimate inventory, which makes the two hard to tell apart from the bid sheet alone. A network that publishes its IVT rate per zone, rather than an aggregate account-level figure, is filtering at the level that actually matters, and the same zone-level disclosure is worth checking against a second source such as push ad network before a budget gets committed.

None of this makes the format obsolete, only smaller and more selective than during the 2016 to 2019 peak of push ads spending. Checking base age, IVT disclosure and carrier-level targeting before committing a budget still returns a margin that display and native both lost years ago.