Last updated: 26 August 2026

What decides whether push ads earn anything at all

The creative matters considerably less on this format than anywhere else across paid media. What decides the outcome is the subscriber database underneath the campaign: where it was collected, how old it is, and how many advertisers already reached it this week. Push ads sold as a single product are never actually a single product, because a list gathered nine days ago and one gathered two years ago behave like separate channels at separate prices. Ask for the age split before agreeing any bid, and treat a refusal as an answer.

Subscriber lists that sit underneath push ads

A network selling push ads sells access to people who once clicked a browser prompt, and everything that follows depends entirely on the way that click was obtained. Prompts placed on a content page produce subscribers who genuinely wanted something from that page.

Incentivised prompts, where a download or a video arrived only after subscription, produce volume and almost no intent whatsoever. Co-registration lists sit lower still, and most advertising platforms resell them without saying so. They arrive cheap because somebody monetised that attention twice already, and the useful clicks went to the first buyer. Every network will still call such a pool fresh if the question is put loosely enough.

How a list ages after collection

Decay is the number nobody prints on a rate card. A base loses subscribers steadily as people clear browser data, switch devices or revoke the permission entirely, and the ones who stay grow slower to respond with every additional campaign they are shown that week. The opening week after collection is worth several times the third month on identical creative, which is why segment age belongs in the bid itself rather than in a footnote on the invoice. Nobody volunteers that split, though every network can produce it when a buyer asks directly and declines to launch without seeing it first.

Networks letting you buy by age band describe something real, and the price gaps between those bands say plainly what they already know about their own supply. Ones selling a single undifferentiated pool have already decided for you, and the blend that results rarely favours whoever is paying.

Creative limits that shape push ads

Character caps do more damage to push ads than any policy rule, because the operating system truncates without warning while the buyer keeps seeing the full version inside the panel. Around thirty characters survive in a title on most Android builds. That is far fewer than native ads allow in a headline, and the figure moves with the launcher rather than the network. The body runs to roughly forty-five characters before an ellipsis appears, and desktop notifications cut at different points again, which means one creative renders three ways across a single campaign.

Write for the truncated version rather than for the one sitting in the preview pane. If an offer only makes sense once the whole line renders, most of the audience never sees the sense in it, and the click rate reported back looks like a creative problem rather than a display one. Check both states carefully before sending anything off for approval.

Icon, title and body inside the character caps

The icon carries recognition at a size where detail disappears entirely, so a logo with fine lettering reads as a grey square on a phone. Solid shapes and two colours survive the resize; anything requiring a squint does not. A large image beneath the text renders on some platforms and vanishes silently on others, which makes it a bonus rather than a component of the message. Build the creative so it works without one. Add the image afterwards, then treat any lift it produces as a bonus rather than as part of the plan.

Emoji pass moderation on most networks and shift click rate measurably. The direction of that shift depends on the vertical rather than the emoji, so test them as a variable, in the same pairs you would run on popunder ads or anywhere placement outweighs creative.

Policy rejections cluster in a single place, which is text imitating a system message of any kind. Anything reading as a delivery notice, a security alert or an unread message gets refused by every network that reviews creatives at all, and the ones approving it are telling you something. Buyers treating every network as interchangeable discover this on a first submission, usually after building forty variants.

Pricing and pacing for push ads

Most push ads inventory clears on cost per click, which suits the format particularly well because an impression on a locked screen means almost nothing at all. Bids move with segment age far more than with geography. A fresh tier one base costs several times an aged one in the same country. Paying tier one rates for an aged base remains the commonest way to lose money on this format entirely. Volume behaves differently too, because a list delivers most of its response in the first hours after a send, so budgets spend far faster here than on an auction-metered channel.

Pace it deliberately. A campaign emptying its budget by ten bought one time zone and called it a country, so hourly caps matter more than daily ones. Split the budget across the day before splitting by source, as you would pace website traffic anywhere else, because delivery hour moves results further than the source list, because a list responds to when it was reached long before it responds to what it was shown.

Frequency caps and send windows

One notification per subscriber per day is the working ceiling, and networks permitting three are simply selling irritation that the publisher eventually pays for through unsubscribes. The cost lands on somebody eventually, and rarely on the buyer who caused it.

Send windows matter every bit as much as the caps themselves, and they cost far less to test properly. Evening delivery outperforms morning on consumer offers in most markets, while business verticals invert that completely, and the gap between the two is wider than anything a headline change will ever produce. A notification arrives whenever you decide that it arrives, which turns timing into a lever rather than an accident of inventory.

Retargeting works better here than on adult traffic bought by placement, because the subscription itself is a durable identifier that survives every cookie restriction shipped so far. A visitor who subscribed through your own prompt stays reachable months later.

In-page delivery where classic push ads cannot fire

Apple devices never accepted web push in the browser the way Android did, which removed a substantial share of tier one inventory from the format overnight and reshaped what the word coverage means on a media plan. In-page units filled the gap by imitating the notification visually inside the page itself, without touching the notification system at all. They reach an audience that classic push ads cannot, they need no subscription at all, and they behave far more like a display placement in every metric that follows the click, including the ones that disappoint.

Treat the two as separate line items with separate targets and separate creative. Blending the two produces a single average that describes neither format, and the reporting on Adult Web Traffic makes exactly the same point about mixing formats inside one campaign.

Measuring a list rather than a campaign

Cohort the results by subscription date rather than by campaign, since that is the variable that actually moves underneath. A click rate falling week on week usually means the base aged rather than the creative tired, and swapping creative in response wastes a fortnight proving it. Cohort reporting takes an afternoon to set up and then answers that question permanently, for every campaign that follows it afterwards.

I worked through the segment definitions published on push-ads.io while rebuilding a client's reporting sheet, and the useful part was a reminder that revenue per thousand subscribers, rather than per thousand impressions, is the only number here that survives contact with a real budget for long. The same discipline applies to any source that keeps its inventory age off the invoice, which is most of them. Revenue per subscriber travels between networks as well, so a base priced badly in one place gets identified quickly in another, and that comparison costs nothing.

Judge a network on how much of that it will tell you before the deposit clears. Operators who publish age bands and unsubscribe rates are running push ads as a product, while the ones quoting a single blended price are running a clearance sale on somebody else's attention, and the difference shows up in month two rather than week one.