What is push notification traffic? Definition, mechanics, and when push is the wrong format

Push notification traffic is what happens when a browser-level notification you opted into months ago turns out to be an ad rather than a message from your bank. The mechanic is the same; the sender is different. That description is most of what a buyer needs to understand the format. Everything below this paragraph is the structural detail that determines whether push fits your campaign.

My name is Priya. I spent five years at Mobidea working on push attribution, fraud detection, and audience-fatigue modelling. I’m writing this for the audience that lands on a glossary expecting either a vendor sales pitch or a Mailchimp-style definition that doesn’t tell them anything useful. This page is neither. It’s the data behind the format, with sample sizes.

The technical definition

Push notification ads are advertiser-paid messages delivered through the browser-level push notification system — the same system that delivers your email client’s new-mail alerts, your news app’s breaking-news pings, or your bank’s transaction confirmations. The system is implemented in Chrome, Firefox, Edge, and Samsung Internet through the W3C Push API standard. Safari implements its own variant that affiliate-push networks generally don’t address at scale.

The lifecycle is three steps:

1. Subscription: A user visits a publisher site that prompts them to “Allow notifications.” The user clicks Allow. Their browser registers a push subscription with the publisher’s service-worker endpoint.
2. Inventory aggregation: The publisher (or a network they’ve integrated with) batches subscribers into segments — usually by GEO, browser, and acquisition date.
3. Ad delivery: The advertiser bids on impressions against those segments. The network’s server-side push gateway delivers the notification to the user’s device. The notification appears as a small panel with a headline, a one-line body, and an optional icon.

The user sees the notification regardless of which browser tab they have open, regardless of whether their browser is foregrounded. They can click it (resolving to the advertiser’s landing page) or dismiss it. On most platforms the notification lingers for 4–8 seconds before auto-dismissing on desktop, less on mobile.

What “push subscription” actually is, mechanically

The subscription that powers push ads is not a marketing list in the conventional sense. It’s a browser-level trust grant. When the user clicked Allow, the browser registered a unique endpoint URL with the publisher’s service worker. That endpoint can be used to deliver messages to the device until the user revokes the permission (which most users never do because most users forget which sites they subscribed to).

Two practical consequences:

The subscription is cookie-independent. A user clearing cookies, switching browsers, or installing a VPN does not lose their push subscription. This is why push retargeting outperforms display retargeting against cookie-cleared cohorts: my Q2 2024 measurement showed +0.12pp absolute CR on push retargeting against cookie-cleared users, n=1.2M, p<0.01. The format reaches a cohort other channels can’t.

The subscription decays. Subscriptions acquired more than two years ago convert at less than 0.1% CR in my Mobidea dataset, regardless of vertical — roughly 28% of total network inventory falls into this stale category. Networks vary widely in how aggressively they prune. Some never prune, which means “premium” labels on stale subscribers are decorative.

Browser segmentation moves CTR more than GEO moves it

The most underweighted fact in push buying: browser distribution within a GEO drives CTR variance more than the GEO itself does, holding vertical constant.

Chrome desktop subscribers convert at 1.4–1.8x the rate of Edge desktop and 2.1–2.8x the rate of Samsung Internet mobile in my Q3 2024 measurements. Most network panels surface GEO as the primary filter and bury browser two clicks deeper. Buyers who segment by browser first and GEO second outperform — the variance is genuinely that large.

The mechanism: Chrome desktop subscribers tend to be acquired through publisher flows with explicit-consent prompts, on devices where push notifications occupy more screen real estate, in workflows that allow longer attention than mobile contexts. Samsung Internet mobile subscribers are disproportionately acquired through softer opt-in flows on lower-priced devices in markets where push-ad arbitrage is heavier. The aggregate effect is a 2–3x CTR delta that’s structural, not creative.

CTR and CR baselines you can actually use

Here are the numbers from my Mobidea aggregated dataset for the verticals I have density on, all from 2024 unless otherwise noted, all with n>1M impressions.

The pattern is consistent: higher CTR verticals (sweepstakes) have lower CR because the click is more impulsive. Lower CTR verticals (utility) have higher CR because the click reflects a clearer match between user need and offer. The CTR-to-CR ratio is the diagnostic; the absolute CTR is decorative.

The mechanics of running a push campaign

The operational shape of a push campaign differs from search or display in three ways that matter for ROI math.

The auction warms up. The first 3–5 days of any new push campaign is auction warm-up, publisher rotation, and fraud-filter training. CR numbers from days 1–3 are not optimisation data; they’re noise. Decisions made on day-2 data systematically over-fit the early publisher rotation. The day-7 number is the earliest stable read for iGaming and finance verticals.

Audience fatigue collapses around week 3 for single-creative runs. Cohort CR decays at -6% to -10% week-over-week from week 4 onward in my iGaming and sweepstakes datasets. Without rotating creative AND refreshing publisher inventory, a push campaign is net-positive on days 1–18 and net-negative thereafter. The renewal that “just needs another month” almost always doesn’t.

Frequency caps below 3/day kill scale. Pushing the cap from 5/day to 3/day in Tier-1 iGaming reduced impressions 36% and improved CR 0.04pp absolute, n=12.4M, Q1 2024. Pushing from 3/day to 1/day reduced impressions another 47% and improved CR 0.02pp absolute. Below 3/day, you’re paying roughly 4x the impression cost per incremental conversion. The “audience experience” framing is brand-campaign reasoning applied to performance budgets where it doesn’t fit. Decide which you’re running.

When push is the wrong format

This is the section you won’t find on most network sales pages, and it’s the section that matters most for buyer decision-making. Push is the wrong format for any of the following:

B2B SaaS. Decision cycles are 14–90 days. Multiple stakeholders. Push delivers 1–3 seconds of decision window. My Q1 2024 parallel test on a B2B vertical I won’t name: push CR 0.03% (n=440K impressions), Google non-branded CR 1.8% (n=82K clicks). The structural mismatch is the decision window. Don’t force this.

Considered purchases above $500. Mortgage rates, insurance quotes, expensive hardware. The user wants to compare, read reviews, ask a partner. Push doesn’t deliver into the comparison phase; it interrupts it. CR collapses to near-zero. The format doesn’t fail because push is “low quality” — it fails because the offer doesn’t fit the format’s structural window.

Anything requiring more than 5 seconds of evaluation. The notification panel auto-dismisses on a window of seconds. If the offer is “click here to see if you qualify for X” and the qualification logic needs context the headline can’t carry, the user dismisses. The format selects for clicker intent, not researcher intent.

Long-tail attribution windows. If the offer’s natural conversion latency is 30+ days, push attribution is poorly suited even on a 14-day server-side window. My iGaming attribution log shows 17% of deposits on days 8–30; a 30-day-plus offer (mortgage refinance, life insurance) is even more skewed and the format under-credits accordingly.

Strict GEO compliance verticals. Push targeting reliable to country level, inconsistent at sub-country resolution across networks. Regulated iGaming, state-licensed insurance, prescription-restricted health products — the operational risk of serving wrong-region traffic is higher than the format’s economic upside in many of these cases. Google’s geo-targeting is reliable to ZIP-code resolution, which is the better operational choice.

When push is the right format

The mirror image of the anti-fit list. Push fits when:

• The offer is low-consideration with a value prop that fits in a six-word headline (iGaming “free spins,” sweepstakes “spin to win,” VPN “block ads everywhere,” AV “scan your PC”).
• The conversion event happens within 7–14 days of the click (deposits, app installs, lead form submissions, not extended-cycle B2B).
• The unit economics tolerate 8–14% raw inventory fraud that drops to 2–5% post-network-filter and lower with your own server-side validation.
• The audience-reach goal is broad rather than narrow — push works against subscribed pools, not against intent queries.
• The campaign can sustain a 21-day minimum test budget through the auction warm-up and day-7 CR stabilisation phases.

If five of those five are true, push is worth a parallel test against your current channel mix. If three of five are true, run a small first-look. If fewer than three are true, push probably isn’t the right answer; look at native, contextual, or intent-based channels instead.

Compliance: what GDPR and the regulators actually require

EU push advertising operates under the ePrivacy Directive’s consent requirements as clarified by national supervisory authorities. The 2024 CNIL (France) and Garante (Italy) guidance both reaffirmed that push notification subscription requires affirmative consent, that the consent must be revocable, and that “Allow” prompts cannot be bundled with cookie-banner consent without an independent affirmation.

Practical consequences for advertisers buying push inventory in the EU:

• Publisher-side compliance is the publisher’s problem, but advertisers buying segmented push inventory through a network are downstream of the publisher’s consent flow. If the publisher’s flow is non-compliant, the inventory is technically at risk.
• Networks are increasingly contracting compliance representations — the standard insertion order for EU push buys now usually includes a publisher-side consent representation. Read the IO.
• Consent withdrawal mechanisms must be operational. A user who turns off push notifications in their browser should not continue receiving impressions from inventory bought against their subscription. Most networks honor this in practice; the long tail of small publishers is less reliable.

I’m not your lawyer. The above is what I learned in five years of running operations adjacent to compliance. For specific campaigns in regulated verticals, get the compliance posture in writing from the network before spending.

FAQ

Is push notification traffic the same as in-page push?

No, but the formats are converging. Classic push is delivered through the browser’s native notification system after a subscription opt-in. In-page push is a styled notification-like banner served inside the publisher page itself, with no subscription opt-in required. In-page CTR runs 1.2–2.8x classic push CTR in my Q1–Q3 2024 measurements, but CR runs 0.6–0.9x. Adjusted for full-funnel performance, the formats are within ±10% on most Tier-1 GEOs.

How does push attribution work?

The network reports a click event when the user taps the notification. The advertiser’s tracker (Voluum, Bemob, server-side) attributes that click against the conversion event using a window of 1, 7, 14, or 30 days. The window matters: iGaming deposits land 38% on day 0, 26% on days 1–3, 19% on days 4–7, 17% on days 8–30 in my Mobidea aggregated dataset. A 7-day window captures 83% of deposits; a 30-day window captures essentially all. Networks default to shorter windows because the auction settles faster.

How fast does push CTR stabilise?

Days 5–7 for iGaming and finance verticals. Days 7–10 for utility verticals. The first 3–5 days are auction warm-up and publisher rotation, not optimisation data. Buyers making decisions on day-2 metrics are deciding on noise. Wait for the day-7 read.

What is a push subscription worth?

It depends entirely on how recently the subscription was acquired and which browser registered it. A 30-day-old Chrome desktop subscription in Tier-1 EU converts at 5–8x the rate of a 2-year-old Samsung Internet mobile subscription in the same GEO on the same offer. Subscription age is the under-discussed quality lever; browser segmentation is the second. CR-per-subscription as a metric is most useful at the sub-source level, not at the network aggregate.

Is push notification advertising still effective in 2026?

For the verticals it fits (low-consideration, short decision window, cookie-independent reach), yes — CTR and CR baselines have been stable across 2022–2024 in my measurements, with audience-quality pressure pushing networks to prune stale subscribers more aggressively. For the verticals it doesn’t fit (B2B SaaS, considered purchases, geo-licensed restricted verticals), push was never effective and isn’t becoming so. The format’s structural fit hasn’t changed.

Next steps

If you’re trying to figure out whether push fits your campaign, the push vs Google Ads conversion data post walks through the comparison with parallel-network data. If push does fit and you’re looking at networks, the 2026 push network ranking compares ten networks on per-vertical CR, fraud rate, and panel transparency. The terms cover the affiliate-disclosure mechanics on adsy.tech links.

For external context, the W3C Push API specification is the canonical technical reference for how the underlying browser system works, and the CNIL guidance on push notification consent is the regulator-side reference most relevant for EU buys.

Numbers cited are from the aggregated Mobidea dataset (n>120M push impressions across 2019–2024) and parallel-network test buys during my consulting work. All data is de-identified before publication. Corrections welcome: [email protected].