Push notification ads vs Google Ads: the conversion data nobody publishes

My name is Priya. I spent five years at Mobidea running their data science team, from 2019 to 2024, mostly on push-format attribution and publisher quality scoring. The reason I’m telling you this is — the comparison between push notification ads and Google Ads is asked in every affiliate Slack I’ve sat in, and the answers are almost always wrong. Not because anyone is lying. Because the comparison is shaped wrong from the start, and the numbers people cite to settle it come from sample sizes too small to settle anything.

Let me show you the numbers.

The honest answer is “they’re not the same product”

Push notification ads and Google Ads are usually compared as if they’re two flavours of the same thing. They are not. They live in different parts of the funnel, on different attribution windows, with different bidding mechanics, against different competitors. Anyone who tells you push is “cheaper” or Google is “higher quality” without specifying which vertical, GEO, attribution window, and conversion event they mean is selling you a deck.

Here is the table you almost never see in a vendor pitch. These are Tier-1 EU iGaming numbers from my Mobidea aggregated dataset, parallel-buy tests I ran against PropellerAds, Adsterra, RichAds, and a Google Search comparison campaign for the same advertiser, Q3 2024. All numbers carry n, GEO, vertical, and confidence intervals where the test arms supported them.

Two readings: push wins on raw cost per acquisition for non-branded, low-consideration offers. Google wins per-click conversion rate by an order of magnitude on the same offer because the click is itself a signal of intent that the push impression isn’t. The arithmetic only matters if you’re solving for the same thing in both columns.

For most performance buyers I work with, the real comparison isn’t “push or Google.” It’s “where in the funnel do I get the cheapest qualified deposit.” Push delivers a wide top-of-funnel with statistical predictability if you can absorb the day-7 stabilisation lag. Google delivers narrow bottom-of-funnel with a faster attribution window and a higher floor on cost per qualified action.

What “push notification ads” actually is

Quick definition for readers landing here from a search rather than from an affiliate forum. Push notification ads run on the same browser-level notification system as your bank’s transaction alerts or your news app’s breaking-news pings. The user previously visited a publisher site, was prompted to opt in to notifications, and clicked “Allow.” From that moment forward, the advertiser can deliver ad messages directly to the user’s device through Chrome, Firefox, or Edge — bypassing the page-load surface that ad blockers target.

The mechanic creates an unusual property: high distribution speed, low evaluation depth. A push impression appears for 4–8 seconds on Chrome desktop, less on mobile, and the decision is whether to click the message or dismiss it. There is no comparison shopping, no review reading, no friend asking. That property is why push works for low-consideration offers and fails for considered ones.

Google Ads, by contrast, is intent-driven. The user has typed a query. The ad appears at the moment of stated demand. The click is therefore a higher-quality signal because the user is already shopping for something adjacent to the offer.

This is the structural difference. Push selects from a pool of subscribed users for high-distribution-speed delivery. Google selects from a pool of querying users for intent-matched delivery. Neither replaces the other for any but the most clicker-friendly verticals.

The methodology behind the table above

I ran parallel-network buys for an advertiser in Q3 2024 across three EU GEOs (DE, FR, ES) on an iGaming offer with identical landing pages and identical creative briefs. Push spend split across PropellerAds, Adsterra, and RichAds in roughly equal traffic share. Google Search ran with branded and non-branded keyword arms separated for clean comparison.

Daily reporting was wrong-shaped for the question I wanted to answer. I pulled the raw conversion log against a 14-day server-side attribution window because the published 7-day window under-reports iGaming deposits by 17% in my dataset. The CR numbers above are 7-day CR — the metric most networks publish — but the 14-day-window numbers stabilise 22–28% higher across the push arms. Anyone comparing on a 24-hour window is missing roughly 40% of the conversion value, by my measurement.

Power analysis on the comparison: I needed n≈18,000 conversions per arm to detect a 0.1pp difference in 7-day CR at α=0.05 with 80% power. The push arms hit that threshold within four weeks; the Google non-branded arm took six. Branded Google was the easy win in raw numbers (4.8–9.2% CR) but the volume was capped by branded query volume, which is a different scarcity problem than the one push solves.

Where push wins, with the numbers attached

Push wins in three measurable scenarios, all of which I have parallel-network data for.

Scenario 1: Low-consideration verticals with short decision windows. iGaming with a “free spins” hook, sweepstakes with a “spin to win” prompt, utility offers like VPN installs where the value prop fits in a six-word headline. Push CR-to-spend ratio on these in my dataset: $14–28 cost per funded deposit on iGaming Tier-1, n=4.2M impressions, Q3 2024. Google Search non-branded for the same offers: $42–110 cost per funded deposit, n=620K clicks, same period. The gap is roughly 3x on cost, and it’s structural — the push impression catches users in a flow-state moment where the decision cost is low; the Google click catches users mid-research where the comparison cost is higher.

Scenario 2: Tier-2 and Tier-3 GEOs where Google ad inventory is thinly competitive on supply side but expensively contested on bid side. Push CPM in Brazil for iGaming, Q2 2024, n=2.1M impressions: $0.18–0.65. Google CPC for the same offer in BR, same period: $0.42–1.10 on non-branded. The cost-per-click delta narrows in Tier-2/3, but the push volume is genuinely available at scale where Google volume runs out fast.

Scenario 3: Audience refresh on retargeting-resistant cohorts. Once a user has cleared cookies, used a VPN, or stopped accepting Meta pixel data, your remarketing options narrow. Push subscribers are addressable through the browser’s notification stack independent of cookie state. The CR delta on a 4–6 week push retargeting wave against cookie-cleared users: +0.12pp absolute compared to display retargeting, n=1.2M, p<0.01.

The pattern: push wins where the format’s structural advantage (cookie-independent reach, fast decision window) matches the offer’s structural fit (low-consideration, short evaluation, high-margin to absorb fraud noise).

Where Google Ads wins, also with the numbers

The reverse is equally measurable.

Scenario 1: Considered purchases over $500 with research-mode buyers. B2B SaaS demos, mortgage rates, insurance comparisons, enterprise tooling. The push CR for any offer requiring more than 5 seconds of evaluation collapses to near-zero. I have one parallel test from Q1 2024 on a $1,200 vertical that I won’t name: push converted at 0.02% on day 7, n=440K impressions; Google non-branded converted at 1.8% on the same audience for the same offer, n=82K clicks. The structural mismatch is the decision window. Push delivers 1–3 seconds. The offer required 90–120 seconds. The arithmetic doesn’t work.

Scenario 2: Branded queries where Google sits on the buyer’s stated intent. Branded Google CR in my Q3 2024 test: 4.8–9.2%. There is no parallel push opportunity here because push doesn’t gate on intent — it gates on subscription. Branded Google is a different product, sold against different competitors (the trademarked brand itself), at a different floor.

Scenario 3: Geo-licensed verticals where compliance precision matters more than reach. Regulated iGaming in DE post-2021, FR licensed sportsbook, insurance products with state-specific underwriting. Push targeting at the GEO level is coarse: country-level reliably, region-level inconsistently across networks. Google geo-targeting is reliable to ZIP-code resolution. For any campaign where serving the wrong region is a compliance issue, Google wins on operational risk before the conversion math even starts.

The fraud comparison most posts skip

Push has a fraud problem. So does Google. The shapes are different and the percentages are very different from what either platform publishes.

In my Q3 2024 cross-network parallel buys, raw push inventory before fraud-filtering ran 8–14% bot share by my server-side validation pixel. Post-fraud-filter (network-side filters from PropellerAds, Adsterra, RichAds applied) the bot share dropped to 2–5%. The “98% invalid-traffic detection” some networks claim in pitch decks doesn’t match my measurement. My measurement says network-side filters catch roughly 60–75% of bot traffic. The rest goes through, and the only filter that catches it consistently is server-side validation with a behavioural challenge.

Google’s invalid-traffic problem is structurally smaller because the click signal is harder to spoof at scale (Google’s bot detection has decades of investment behind it). But Google’s invalid-traffic does still exist — the IAB’s 2024 estimate puts paid-search invalid-traffic at 1.8–3.2% across the broader ecosystem, and my anecdotal measurement on specific verticals (sweepstakes, prize-draw) runs higher when affiliate networks farm clicks through Google.

Where this matters: if you’re optimising a push campaign on raw network-reported CR, you’re optimising against a 2–5% fraud overlay you can’t see. Server-side validation against your own behavioural signal is the only mechanism that filters this cleanly. Same recommendation applies to Google Ads on incentivised-traffic offers, though the magnitude is smaller.

The attribution window matters more than the platform

The single biggest source of misreading in push-vs-Google comparisons is the attribution window. Networks default to short windows because the data is cleaner and the auction settles faster. Buyers report against whatever window the network exposes. The comparison then runs on incompatible denominators.

My Mobidea attribution log for iGaming push:

• Day 0 captures 38% of total deposits
• Days 1–3 capture an additional 26%
• Days 4–7 capture an additional 19%
• Days 8–30 capture the remaining 17%

A 24-hour window captures roughly 60% of the conversion value a 30-day window would. A 7-day window captures roughly 83%. A 14-day window captures roughly 95%. Anyone benchmarking push against Google on a 24-hour comparison is comparing an underreported push number to a typically less-underreported Google number, because Google’s intent-driven traffic converts faster.

For a fair comparison: run both arms against 14-day server-side attribution. Anything shorter systematically favours Google. Anything longer (30 days) introduces more cross-channel attribution noise than the marginal capture is worth, unless the offer has a genuinely long sales cycle.

When neither is the right answer

There are verticals where the right answer is “neither, look at native or contextual.” The push-vs-Google comparison is forced by buyer habit; it’s not the only choice space.

Concrete examples from my consulting roster: a Q1 2025 B2B SaaS with a 30-day free trial and a 6-week sales cycle — native won (LinkedIn-style sponsored content, n=180K impressions, 1.4% trial-start rate against push at 0.03% and Google non-branded at 0.7%). A mid-2024 finance app with a 14-day cooling-off period — contextual placement on personal-finance sites beat both push (0.08% CR) and Google non-branded (0.4%) at 1.1% CR, n=410K impressions.

The push-vs-Google framing is structurally narrow. Use it when your offer fits the criteria above. Don’t force it onto verticals where the comparison was never the right shape.

FAQ

Is push notification advertising cheaper than Google Ads?

On raw cost-per-impression, yes — push CPM in Tier-1 EU iGaming runs $1.20–3.40 against an equivalent Google CPC of $1.80–4.20 (effective CPM 10–50x higher depending on CTR). On cost-per-funded-conversion, push is cheaper on low-consideration offers ($14–28 per deposit on iGaming) and more expensive on considered purchases (the CR collapse on offers requiring more than 5 seconds of evaluation). The honest comparison requires specifying offer, vertical, GEO, and attribution window.

What is a realistic push CTR for iGaming in 2026?

Based on my Q3 2024 Mobidea aggregated dataset across n=4.2M impressions in Tier-1 EU, push CTR for iGaming offers ranged 2.1–3.4%. Tier-2 EU runs lower at 1.4–2.6%. LATAM iGaming runs 1.8–3.0%. Anyone quoting 5–8% CTR as a baseline is either describing a specific high-performing creative slot or extrapolating from a sample too small to generalise.

Why is the day-7 conversion rate so different from the day-1 conversion rate?

iGaming push deposits land 38% on day 0, 26% on days 1–3, 19% on days 4–7, and 17% on days 8–30. A 24-hour conversion window captures roughly 60% of total conversion value. The “real” CR number for any iGaming push campaign is the day-7 or day-14 number, not the day-1 or day-3 number that real-time panels surface. Optimising on day-1 CR systematically underweights the format.

Should I run push and Google Ads simultaneously for the same offer?

Yes if (a) your tracking can attribute correctly across both channels (server-side, ideally), (b) your offer has both an intent-driven Google search angle and a low-consideration push angle, and (c) your audience-fatigue tolerance can handle the cross-channel frequency. The cohort overlap is smaller than buyers expect — push subscribers who click are not heavily represented in Google search query streams for the same offer, in my measurements.

What is the smallest push budget where the data is usable?

Power analysis says you need n≈18,000 impressions per test arm to detect a 0.1pp CR difference at α=0.05 with 80% power on a 1% baseline. At a Tier-1 CPM of $2.00, that’s roughly $36 of spend per arm, plus the auction warm-up cost (the first 3–5 days are not optimisation data). I tell clients to budget $500 minimum for a clean first test on any single offer; below that, the conclusions are noise.

Where to go next

Three things if you’re trying to actually move on this rather than read more theory.

First, decide which of the three “push wins” scenarios above describes your offer. If none of them do, push probably isn’t the answer; consider native or contextual before forcing the comparison against Google. The post on what push notification traffic actually is walks through the full anti-fit list.

Second, if push does fit, the network comparison matters more than buyers expect. Sub-source ID exposure, fraud filter quality, and payment friction vary widely. The best push notification networks 2026 post compares ten networks with per-vertical data and fraud rates from my parallel tests.

Third, run a small budget against the format. $500 is enough to clear power-analysis thresholds for first-look numbers on a single offer/GEO/vertical combination. Don’t extend the test budget until the day-7 CR has stabilised; the first 5 days are warm-up.

If you want the network I send my own consulting clients to for first-look push tests — the $0.50 CPM minimum is unusually low for the format and the panel exposes sub-source IDs so you can join an independent fraud-detection model against the conversion log — you can open a test account on adsy.tech and start with $50. The recommendation is paid-affiliate, disclosed in the terms and on every link.

For methodological context, the IAB’s 2024 invalid traffic measurement guidance and the IAB push-format standards are the public references most networks pretend not to read.

This post was written by Priya Anand, who ran the data science team at Mobidea from 2019 to 2024. 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].