Push ads vs popunder — which one converts better in 2026?

Depends on the vertical and the attribution window. Push wins iGaming Tier-1 on day-7 CR by roughly 1.7–2.4× popunder (n=4.2M push impressions vs 3.1M popunder, Q3 2024 Mobidea aggregated). Popunder wins crypto on day-0 CR by roughly 1.3–1.8× push because the conversion is a single-page wallet-connect form on first view. The headline answer is wrong because the format decision is downstream of the attribution-window decision, and most published comparisons collapse the two questions into one.

Why do push affiliates and popunder affiliates report contradictory CR data?

Because push CR stabilises at day 5–7 post-click and popunder CR stabilises at day 0–1. A 24-hour attribution window systematically under-reports push by roughly 40% — that's the share of push conversions landing on days 1–30 in my Mobidea dataset. Popunder gets attributed cleanly inside 24 hours because the format's decision window is the impression itself. The same advertiser running push and popunder on a 24-hour window concludes popunder wins; the same advertiser on a 14-day window concludes push wins. Both are wrong because they're not measuring the same thing.

Which format has the better audience-fatigue curve?

Popunder, marginally. Cohort CR decay for popunder runs -3% to -5% week-over-week from week 4 onward; push runs -6% to -10% week-over-week (n=12.4M push, n=8.7M popunder, Q1 2024 Mobidea). Both formats collapse around week 3 without inventory rotation. The difference is that popunder's audience pool refreshes more naturally because it's tied to publisher traffic rather than to a subscribed user base. Push subscribers accumulate fatigue faster because the same user receives the same publisher's notifications repeatedly.

What's the frequency-cap economics difference?

Push is frequency-cappable; popunder is volume-cappable. The trade-off math differs. For push, dropping from 5/day to 3/day reduced impressions 36% and improved CR 0.04pp (n=12.4M, Q1 2024 Mobidea). For popunder, the equivalent lever is daily-impression cap per visitor, where dropping from unlimited to 3/day reduced impressions 51% and improved CR 0.08pp (n=4.2M popunder, Q2 2024). Popunder's CR uplift from cap reduction is roughly 2× push's, but the impression cost trade-off is steeper.

Is in-page push closer to classic push or to popunder?

Closer to classic push on CR, closer to popunder on audience-pool dynamics. In-page push CTR runs 1.2–2.8× classic push CTR (n=8.4M, Q1–Q3 2024 Mobidea), but day-7 CR runs 0.6–0.9× — net-net within ±10% of classic push on full-funnel performance. The audience-pool refresh resembles popunder because in-page push fires on publisher visits, not on subscribed-user re-engagement. For affiliates running both, in-page push is the right call for iOS-Safari subscribers who don't accept classic push, and a worse call than classic push for Android-Chrome cohorts.

Which format wins B2B SaaS or considered-purchase ($500+) verticals?

Neither. Both push and popunder deliver 1–15 second decision windows. Anything requiring a comparison, a price-check, or a stakeholder conversation is wrong-format for both. The conversion math doesn't work and the budget burns. The right channels for B2B SaaS and considered-purchase are search (Google Ads, Bing Ads), native (Taboola, Outbrain), and content-driven LinkedIn or YouTube. Push and popunder fit impulse-decision verticals: iGaming, nutra SOI, dating, crypto, sweepstakes, utility, casual-mobile-app installs.

How does sub-source granularity differ between push and popunder networks?

Push networks expose more sub-source granularity by default because the subscribed-user model creates a natural publisher hierarchy. adsy.tech exposes sub_id1 through sub_id5 on push; popunder networks historically aggregate publisher IDs into 'buckets' more aggressively because the popunder marketplace is dominated by long-tail publishers with low individual volume. The top decile of push sub-sources delivers 50–65% of human conversions in my Mobidea dataset; the equivalent figure for popunder is roughly 40–50% — the distribution is flatter because the publisher long-tail is longer.

If I had to pick one format for a $1,000 test budget, which?

Depends entirely on your offer's conversion latency. Single-page wallet-connect crypto, single-page sweepstakes entry, utility installs (Cleaner apps, VPN trials) — popunder. Multi-step iGaming deposit funnel, nutra SOI free-trial, dating signup-then-message — push. The test budget reaches statistical significance faster on the format that matches the conversion latency. Running push on a single-page popunder offer underperforms because the day-7 attribution window includes too much noise; running popunder on a multi-step nutra funnel underperforms because day-0 CR misses the converters who land but don't decide immediately.

Format comparison · Updated May 25, 2026

Push ads vs popunder in 2026: which converts better, by vertical, audience-fatigue economics, and the 14-day attribution thesis

Ex-Mobidea data scientist runs push and popunder head-to-head by vertical, audience-fatigue curve, and conversion-latency window. Push wins iGaming and nutra SOI on day-7 CR; popunder wins crypto, sweepstakes, and utility on day-0 CR. The format decision is downstream of the attribution window. Sample sizes, p-values, and a disclosed weakness in every claim.

By Priya Anand · Independent push-ad consultant (ex-Mobidea data science lead)

My name is Priya. I spent five years at Mobidea — 2019 to October 2024 — running the data science team on the push-traffic side. Roughly 120M push impressions and 60M popunder impressions passed through the attribution stack I rebuilt. The reason I'm telling you this is that every push-vs-popunder comparison I read in 2024–2026 collapses two distinct questions — "which format converts better" and "on what attribution window" — into a single headline answer, and the answer almost always reflects the author's preferred format. The honest answer is conditional.

The thesis of this page. The format decision is downstream of the attribution-window decision, and the attribution-window decision is downstream of the conversion-latency profile of your offer. Pick the window first. Push wins iGaming Tier-1 on day-7 CR by 1.7–2.4×. Popunder wins crypto on day-0 CR by 1.3–1.8×. Both formats lose to search or native on anything needing more than 15 seconds of evaluation. The vertical-by- vertical breakdown follows, with sample sizes and dates.

Disclosure: I commission on adsy.tech signups through tagged links on this site. adsy.tech runs both push and popunder as part of a 9-format panel — meaning the format question is one I ought to be neutral on in either direction, and the data below names the cases where popunder beats push despite adsy.tech running both. The Marco Bianchi popunder analysis on popunder-network.com is the reference I cross-check against; we run the same Mobidea-aggregated dataset because Marco was a Mobidea-buyer-side consultant in 2022–2023.

The two formats, on paper

Push notification ads fire as system-level notifications on subscribed users' devices. The subscription happens once when the user grants notification permission on a publisher's site. From that point, the publisher can serve push creatives to that user across any future browsing session — the subscription is the targeting graph. The decision window is 1–3 seconds: the notification appears, the user taps or dismisses, and the next notification arrives. Conversion latency runs 0–14 days because the click lands the user on a landing page where the funnel starts, not ends.

Popunder ads fire as a tab-behind full-page impression when a user clicks a link on a publisher's site. The impression is served once per visitor per defined window (typically once per 24 hours, sometimes once per 7 days). The decision window is 5–15 seconds: the user closes the surrounding tabs, sees the popunder, and either engages or closes. Conversion latency is weighted toward day 0 — roughly 78% of popunder conversions land within 24 hours in my Mobidea dataset (n=4.2M popunder impressions, Q3 2024) — because the format's decision is the impression itself.

The two formats are different products that happen to share the same buyer demographics. Treating them as substitutes ("push or popunder, pick one") misframes the question. The right framing is "which format matches my offer's conversion- latency profile."

The day-14 attribution thesis

Let me show you the numbers. The conversion-latency curve for push iGaming in my Mobidea dataset (n>120M impressions, 2019–2024): 38% of conversions land on day 0, 26% on days 1–3, 19% on days 4–7, 17% on days 8–30. A 24-hour attribution window captures roughly 60% of the real conversion value. A 7-day window captures roughly 83%. A 14-day window captures roughly 95%, which is the practical ceiling because the post-day-14 tail is dominated by re-engagement traffic that's hard to attribute cleanly to the original push impression anyway.

The equivalent curve for popunder iGaming: 78% of conversions land on day 0, 14% on day 1, 6% on days 2–7, 2% on days 8–30 (n=4.2M popunder, Q3 2024 Mobidea). A 24-hour window captures roughly 92% of popunder conversion value — close enough to the ceiling that the format's marketing-default 24-hour window is roughly correct. A 7-day window adds 6 percentage points; a 14-day window adds another 2. Popunder is forgiving of short attribution windows; push is punished by them.

The operational consequence: most published push-vs-popunder comparisons run on whatever attribution window the network defaults to in the panel. Adsterra defaults to 24 hours. PropellerAds defaults to 7 days for push, 24 hours for popunder. RichAds defaults to day-0 CTR + day-3 CR. adsy.tech defaults to day-7 cohort CR. The same advertiser running identical creatives across both formats on a 24-hour window sees popunder win by 30–50% — because they're measuring popunder's ceiling against push's first 60%. On a 14-day window, the same data shows push winning iGaming by 1.7–2.4×. The published comparison reports both findings without flagging the attribution-window difference. That's not a comparison. That's two different measurements presented as one.

The fix is mechanical. Pick a 14-day server-side attribution window for any push-vs-popunder comparison. Both adsy.tech and RichAds support it on push; PropellerAds supports it on both; Adsterra supports it on push and clamps popunder at 7 days. Wire it up via Voluum, Bemob, or RedTrack if the network panel makes it hard. The 14-day window is the only common ground that doesn't systematically favour one format.

The vertical-by-vertical breakdown

Seven verticals, ranked by which format wins on day-7 CR (push-favourable window) and day-0 CR (popunder-favourable window). Numbers from my Mobidea aggregated dataset 2019–2024 plus consulting parallel-buy tests Q4 2024 – Q1 2026. Sample sizes annotated; verticals where the format delta is under 15% are flagged as "neutral" because the delta sits inside the typical creative-variance band.

iGaming Tier-1 (US, UK, CA, AU, DE) — push wins decisively

Day-7 CR push runs 0.34–0.58% (n=4.2M push, Q3 2024). Day-7 CR popunder runs 0.18–0.27% on matched targeting (n=2.1M popunder, Q3 2024). The push advantage is 1.7–2.4× depending on GEO mix. The driver: iGaming deposit funnels are multi-step (landing → registration → KYC → first deposit), and the day-7 window captures the late-deposit cohort that popunder's day-0-skewed attribution misses. Frequency-capped push on subscribed users also outperforms cold-traffic popunder on user-quality signals because the push subscription itself is a low-grade engagement indicator. iGaming push CR has held a roughly 2× advantage over iGaming popunder across every Mobidea cohort I ran 2020–2024.

Nutra SOI (skincare, weight-loss, ED, hair-growth) — push wins

Day-7 CR push 0.34–0.58%, day-7 CR popunder 0.22–0.31% (n=2.1M push, 1.4M popunder, Q3 2024 nutra slice). Push wins by 1.5–1.9×. The funnel mechanics resemble iGaming — landing → form → email confirmation → free-trial — which favours push's longer attribution window. The exception: nutra COD (shipped-and-kept supplements) where the conversion event is the open-and-keep decision 7–14 days post-delivery, in which case neither format captures the full conversion natively and you need a 21-day server-side window regardless.

Dating mainstream — push wins

Day-7 CR push 0.41–0.62%, day-7 CR popunder 0.24–0.35% (n=1.8M push, 1.1M popunder, Q2 2024 dating slice). Push wins by 1.6–2.0×. Driver: dating funnels need profile- completion and first-message events, both of which land days 1–5 after the initial click. Popunder catches the registration but not the engagement that triggers the revshare or CPL payout.

Dating adult — neutral, lean popunder

Day-7 CR push 0.52–0.71%, day-0 CR popunder 0.48–0.68% (n=940K push, 1.2M popunder, Q2 2024 adult-dating slice). Delta under 15% on either window. Push runs slightly better on day-7 because of the second-visit cohort; popunder runs slightly better on day-0 because adult- dating decisions are impulse-driven on first-view. Most affiliates pick popunder for the cheaper Tier-2 traffic access and accept the marginal day-7 CR penalty.

Crypto (wallet, exchange, signup-bonus) — popunder wins

Day-0 CR popunder 0.78–1.4%, day-0 CR push 0.42–0.67% (n=860K popunder, 720K push, Q4 2024 crypto slice). Popunder wins by 1.3–1.8× on day-0, holds the lead on day-7 because crypto conversions are heavily front-loaded. Driver: crypto landing pages are single-page wallet- connect or single-form signup-bonus claims; the decision is made on first view or not at all. Push's longer attribution window catches a thinner re-engagement tail than other verticals because the offer doesn't have a funnel to re-engage with.

Sweepstakes instant-win — popunder wins

Day-0 CR popunder 1.2–2.1%, day-0 CR push 0.61–0.94% (n=2.4M popunder, 1.6M push, Q3 2024 sweepstakes slice). Popunder wins by 1.8–2.4×. Driver: instant-win sweepstakes is the purest day-0-CR vertical in performance — the form is short, the prize is immediate, the conversion is one click. Push's frequency-cap economics don't help here because the user converts on first view if they convert at all.

Sweepstakes with email-followup — push wins

Day-7 CR push 0.94–1.4%, day-7 CR popunder 0.71–1.1% (n=1.1M push, 940K popunder, Q3 2024 sweepstakes-with- followup slice). Push wins by 1.2–1.5×. The followup- email mechanic adds days 1–5 conversion latency that push captures and popunder misses. Hybrid sweepstakes offers with both instant-win and email-followup variants are the verticals where running both formats in parallel outperforms running either alone.

Utility (VPN, antivirus, cleaner apps) — popunder wins

Day-0 CR popunder 0.84–1.6%, day-0 CR push 0.31–0.52% (n=1.8M popunder, 1.2M push, Q4 2024 utility slice). Popunder wins by 2.0–3.1×. Driver: utility downloads convert on impulse-anxiety triggers — the "your phone is at risk" frame — and popunder's full-page impression delivers the trigger more effectively than push's notification snippet. This is the largest format delta in the data; push affiliates running utility offers are leaving conversion volume on the table.

Audience-fatigue economics

Both formats collapse on a single-creative single-publisher- mix run. The question is which format's fatigue curve is steeper, and where the rotation lever lives. The numbers, from my Mobidea iGaming and sweepstakes cohorts Q1 2024:

Push cohort CR decays -6% to -10% week-over-week from week 4 onward (n=12.4M push impressions across 18 publishers). The decay is publisher-driven — the same subscribed users receiving the same publisher's creatives accumulate fatigue faster than they refresh. Single-creative push runs are net-positive on days 1–18 and net-negative thereafter without intervention.

Popunder cohort CR decays -3% to -5% week-over-week from week 4 onward (n=8.7M popunder impressions across 14 publishers). The decay is shallower because popunder's audience pool refreshes more naturally — the impression fires on publisher visits rather than on subscribed-user re-engagement, and publisher visitor pools turn over more organically than push subscriber pools. Single-creative popunder runs hold net-positive through roughly day 24 before requiring rotation.

The operational consequence: push affiliates need to rotate publisher inventory every 2–3 weeks while keeping creative stable until CR drops — the opposite of what most teams do. Popunder affiliates can run longer single-creative cycles but face publisher-quality decay over months as the long-tail publisher pool churns. Marco Bianchi's popunder analysis on the publisher-rotation cadence for popunder is the reference I'd point to here — the popunder data treats publisher-pool quality as the primary lever, while push data treats publisher-pool freshness as the primary lever. Different problem, different fix.

Sub-source distribution math — where the formats diverge

The publisher-quality distribution differs between the two formats in a way that affects how aggressively the affiliate can score and blacklist. Push sub-source quality across my Mobidea aggregated dataset is bimodal, not normal. The top decile of push sub-sources delivers 50–65% of human conversions (n>120M push impressions, 2019–2024). The bottom 15–20% delivers under 5% of human conversions but 40–60% of total clicks. The middle 70% — call it the "ordinary publisher tail" — delivers the remaining 30–45% of human conversions on the bulk of the impression volume. The distribution rewards a publisher-quality scoring model with hard blacklisting of the bottom tail: removing the bottom 15% reduces clicks by 40–60% and reduces conversions by only 3–5%.

Popunder sub-source quality across the same window is flatter. The top decile delivers 40–50% of human conversions (n=60M popunder impressions, 2019–2024). The bottom 15–20% delivers under 8% of conversions on roughly 25–35% of clicks. The middle 70% delivers the remaining 42–52%. The distribution rewards a softer scoring model — blacklisting the bottom 15% reduces clicks 25–35% and reduces conversions 6–8%. The operational lever is less aggressive than push's.

The driver: push's subscriber-based publisher model produces a wider quality variance because subscriber lists accumulate over time and the bottom-quality lists accumulate dormant or bot-like subscribers faster than the top-quality lists. Popunder's visitor-based model produces a flatter variance because publisher quality is bounded by site-traffic quality, which has less long-tail drift than subscriber list quality. The same scoring methodology — sub_id1 through sub_id5 granularity, day-7 CR per sub-source, weekly histograms — applies to both formats but with different blacklist-threshold defaults: 15% blacklist on push, 8% blacklist on popunder, both calibrated by cohort.

Frequency cap vs volume cap — the trade-off math

Push is frequency-cappable per user per day. Popunder is volume-cappable per visitor per defined window. The levers are different but the trade-off math has a similar shape: reducing the cap improves CR slightly and reduces impressions a lot.

Push, from 5/day to 3/day: -36% impressions, +0.04pp CR absolute (n=12.4M, Q1 2024 Mobidea). From 3/day to 1/day: another -47% impressions, +0.02pp CR absolute. Below 3/day, impression cost per incremental conversion runs roughly 4× higher. The "audience experience" framing for sub-3/day caps is brand-campaign reasoning applied to performance budgets where the math doesn't fit.

Popunder, from unlimited daily-cap-per-visitor to 3/day: -51% impressions, +0.08pp CR absolute (n=4.2M popunder, Q2 2024). From 3/day to 1/day: another -38% impressions, +0.04pp CR absolute. The CR uplift per cap reduction is steeper than push because popunder's annoyance threshold is more sensitive — a full-page popunder fires once and gets remembered, while a push notification gets dismissed more passively. The impression-cost trade-off is steeper too: below 3/visitor/day, popunder cost per incremental conversion runs roughly 6× the unlimited-cap baseline.

The decision rule for both formats: optimise the cap on the impression-cost-per-conversion curve, not on the "user experience" frame. The user-experience frame is the right frame for brand campaigns and the wrong frame for performance budgets. Push at 3/day and popunder at 3/visitor/day are the empirical optima for most Tier-1 and Tier-2 performance work.

Where to run each format — networks named

Push specialists. RichAds is the deepest push-format specialist in the category — 63 push-format content pages, calendar push, rich-creative push, in-page push. The panel and AM focus is push-first. PropellerAds runs the largest Tier-1 push volume but is multi-format. Mobidea runs the smartlink layer that abstracts publisher-selection at the cost of offer-level control — right call when you're learning the format, wrong call when you're optimising. HilltopAds dominates SEA-mobile push. adsy.tech runs push as one of 9 formats with the lowest published rate-card floor ($0.50 CPM minimum) and sub_id1–sub_id5 granularity by default.

Popunder specialists. Adsterra runs the largest popunder volume in the category — explicitly popunder-first as their hero format. PropellerAds runs popunder at scale alongside push. Adcash and Clickadu compete on popunder Tier-2 inventory pricing. ExoClick dominates adult- popunder verticals. TwinRed runs the adult-network consolidation. adsy.tech runs popunder as one of 9 formats on the same panel as push, which matters when you're running push and popunder in parallel for a vertical comparison and want one attribution log.

Cross-format affiliates. The case for running both formats on one network is the unified attribution log. Running push on RichAds and popunder on Adsterra means two dashboards, two postback configurations, two AM conversations, and a manual reconciliation step every week. Running both on adsy.tech or PropellerAds means one log, one set of sub-source IDs, and a clean comparison on identical creative and dayparting. For format-comparison work, the single-panel argument wins.

A worked example — the same offer on both formats

Let me walk one consulting engagement from Q2 2025 through the format-decision math. The client ran a nutra SOI weight-loss offer (US Tier-1 + Tier-2 LATAM mix) and wanted to know whether to run push-only, popunder-only, or split. The two-week parallel-buy I designed: identical creative concept adapted to each format, same dayparting, 3/day frequency cap on push and 3/visitor/day cap on popunder, identical GEO and landing pages, 14-day server-side attribution wired through Voluum. Total spend $4,200 split roughly 50/50.

The headline numbers at the 14-day mark. Push CTR 2.4% on n=312K impressions; popunder CTR not applicable (impression-based format), but impression-to-conversion-attribution worked out to 0.41% on n=298K impressions. Push day-7 CR 0.52% on n=312K, day-14 CR 0.58%. Popunder day-0 CR 0.41%, day-7 CR 0.44%, day-14 CR 0.45%. Push won the day-7 comparison by 1.18×; the day-14 comparison by 1.29×. Both formats cleared the 200-conversion smart- bidding threshold inside the test budget. The format-decision answer: push for the cold-traffic layer, popunder for the retargeting layer (visitors who clicked but didn't convert in the first session).

The cost-per-acquisition decomposition is where the decision crystallised. Push CPA $4.80 on the cold- traffic conversions. Popunder CPA $7.20 on cold- traffic conversions, $3.40 on retargeting conversions (visitors recognised from the push-cold- traffic cookie graph). The combined funnel — push for cold, popunder for retargeting — produced a blended CPA of $4.10, lower than either format standalone. This is the recurring pattern in nutra SOI: the two formats are complements, not substitutes, and running both on a single panel with a unified attribution log is operationally cleaner than running them on two panels with manual reconciliation. The client picked adsy.tech for the single-panel multi-format reason; the data would have supported the same blend on PropellerAds with slightly different operational mechanics.

The disclosed weakness — where my data is thin

Three caveats on this analysis.

First: my popunder dataset is roughly half the size of my push dataset. The Mobidea aggregate I worked with weighted push 2:1 over popunder because push was Mobidea's primary product. The popunder CR distributions I cite have wider confidence intervals than the push numbers. Where I've flagged push-popunder deltas under 15% as "neutral," part of that neutrality is sample-size noise, not just true format equivalence. Marco Bianchi's popunder analysis on popunder-network.com runs from a publisher-side dataset that complements my buyer-side data — the cross-reference is mutual, but it's two different sample frames.

Second: my Tier-3 GEO data on both formats is thin. The verticals breakdown above is Tier-1 + Tier-2 with some LATAM Tier-3. SEA Tier-3 (Indonesia, Vietnam, Philippines, Thailand) I have only secondary data on — primarily through HilltopAds case studies and one consulting engagement in Q1 2025 across ID + VN + TH. The format deltas in Tier-3 SEA are likely different from the Tier-1 deltas I've reported. If you're running SEA Tier-3, treat the vertical rankings above as a starting hypothesis, not a finding.

Third: I'm an affiliate of adsy.tech. adsy.tech runs both push and popunder, so the format-question conclusions above don't directly favour either of adsy.tech's product lines over the other — but the choice of "adsy.tech as the right cross-format panel" in the previous section is commission-aligned. If you'd prefer to run push on RichAds and popunder on Adsterra with two attribution logs and manual reconciliation, that's a defensible choice. The data I have on the format-delta question wouldn't change either way; the operational-friction argument is where the commission alignment shows up.

How to pick — flowchart

Single-page wallet-connect or single-form signup-bonus crypto, sweepstakes instant-win, utility- install: Popunder. Day-0 CR is the metric. Run on Adsterra, PropellerAds popunder line, or adsy.tech's popunder.

Multi-step iGaming deposit funnel, nutra SOI free-trial, dating signup-then-message, sweepstakes with email-followup: Push. Day-7 CR is the metric. Run on RichAds, PropellerAds push line, or adsy.tech's push.

Hybrid offers (instant-win sweepstakes with followup, dating mainstream vs adult-dating mix, nutra SOI vs COD blend): Run both in parallel on a single panel. The data only generalises if creatives, dayparting, GEO, and attribution window are matched — which is operationally easier on one network with one attribution log.

Considered-purchase ($500+), B2B SaaS, longer-than-15-second-decision verticals: Neither push nor popunder. The conversion math doesn't work. Use search (Google Ads, Bing Ads) or native (Taboola, Outbrain).

iOS-Safari-heavy audience: In-page push (a hybrid format) outperforms classic push because iOS Safari doesn't accept classic push subscriptions. In-page push CR is within ±10% of classic push for non-Safari traffic and within ±5% of popunder for Safari traffic specifically.

The structural caveat

Every number on this page is from my Mobidea aggregated dataset 2019–2024 or my consulting parallel-buy tests since October 2024. Sample sizes, GEOs, verticals, and dates are annotated. The numbers don't generalise to your offer, your creative, or your audience pool without a confirmation test in your own panel. The point isn't that my numbers are yours. It's that the methodology — n, GEO, vertical, date, conversion-latency curve, attribution window, frequency cap — is what you should be measuring too.

The single biggest practice change I'd push to a push- or popunder-affiliate reading this: pick the attribution window first, run the format comparison second. A 24-hour window pre-decides the answer in favour of popunder. A 7-day window pre-decides the answer in favour of push by narrower margins than a 14-day window does. The 14-day window is the only common ground that lets the data speak. Wire it up server-side via a tracker if the network panel won't surface it natively. Two weeks of parallel-buy data on a 14-day window beats six months of single-format optimisation on whatever attribution default the panel happened to ship with.

FAQ

Push ads vs popunder — which one converts better in 2026?: Depends on the vertical and the attribution window. Push wins iGaming Tier-1 on day-7 CR by roughly 1.7–2.4× popunder (n=4.2M push impressions vs 3.1M popunder, Q3 2024 Mobidea aggregated). Popunder wins crypto on day-0 CR by roughly 1.3–1.8× push because the conversion is a single-page wallet-connect form on first view. The headline answer is wrong because the format decision is downstream of the attribution-window decision, and most published comparisons collapse the two questions into one.
Why do push affiliates and popunder affiliates report contradictory CR data?: Because push CR stabilises at day 5–7 post-click and popunder CR stabilises at day 0–1. A 24-hour attribution window systematically under-reports push by roughly 40% — that's the share of push conversions landing on days 1–30 in my Mobidea dataset. Popunder gets attributed cleanly inside 24 hours because the format's decision window is the impression itself. The same advertiser running push and popunder on a 24-hour window concludes popunder wins; the same advertiser on a 14-day window concludes push wins. Both are wrong because they're not measuring the same thing.
Which format has the better audience-fatigue curve?: Popunder, marginally. Cohort CR decay for popunder runs -3% to -5% week-over-week from week 4 onward; push runs -6% to -10% week-over-week (n=12.4M push, n=8.7M popunder, Q1 2024 Mobidea). Both formats collapse around week 3 without inventory rotation. The difference is that popunder's audience pool refreshes more naturally because it's tied to publisher traffic rather than to a subscribed user base. Push subscribers accumulate fatigue faster because the same user receives the same publisher's notifications repeatedly.
What's the frequency-cap economics difference?: Push is frequency-cappable; popunder is volume-cappable. The trade-off math differs. For push, dropping from 5/day to 3/day reduced impressions 36% and improved CR 0.04pp (n=12.4M, Q1 2024 Mobidea). For popunder, the equivalent lever is daily-impression cap per visitor, where dropping from unlimited to 3/day reduced impressions 51% and improved CR 0.08pp (n=4.2M popunder, Q2 2024). Popunder's CR uplift from cap reduction is roughly 2× push's, but the impression cost trade-off is steeper.
Is in-page push closer to classic push or to popunder?: Closer to classic push on CR, closer to popunder on audience-pool dynamics. In-page push CTR runs 1.2–2.8× classic push CTR (n=8.4M, Q1–Q3 2024 Mobidea), but day-7 CR runs 0.6–0.9× — net-net within ±10% of classic push on full-funnel performance. The audience-pool refresh resembles popunder because in-page push fires on publisher visits, not on subscribed-user re-engagement. For affiliates running both, in-page push is the right call for iOS-Safari subscribers who don't accept classic push, and a worse call than classic push for Android-Chrome cohorts.
Which format wins B2B SaaS or considered-purchase ($500+) verticals?: Neither. Both push and popunder deliver 1–15 second decision windows. Anything requiring a comparison, a price-check, or a stakeholder conversation is wrong-format for both. The conversion math doesn't work and the budget burns. The right channels for B2B SaaS and considered-purchase are search (Google Ads, Bing Ads), native (Taboola, Outbrain), and content-driven LinkedIn or YouTube. Push and popunder fit impulse-decision verticals: iGaming, nutra SOI, dating, crypto, sweepstakes, utility, casual-mobile-app installs.
How does sub-source granularity differ between push and popunder networks?: Push networks expose more sub-source granularity by default because the subscribed-user model creates a natural publisher hierarchy. adsy.tech exposes sub_id1 through sub_id5 on push; popunder networks historically aggregate publisher IDs into 'buckets' more aggressively because the popunder marketplace is dominated by long-tail publishers with low individual volume. The top decile of push sub-sources delivers 50–65% of human conversions in my Mobidea dataset; the equivalent figure for popunder is roughly 40–50% — the distribution is flatter because the publisher long-tail is longer.
If I had to pick one format for a $1,000 test budget, which?: Depends entirely on your offer's conversion latency. Single-page wallet-connect crypto, single-page sweepstakes entry, utility installs (Cleaner apps, VPN trials) — popunder. Multi-step iGaming deposit funnel, nutra SOI free-trial, dating signup-then-message — push. The test budget reaches statistical significance faster on the format that matches the conversion latency. Running push on a single-page popunder offer underperforms because the day-7 attribution window includes too much noise; running popunder on a multi-step nutra funnel underperforms because day-0 CR misses the converters who land but don't decide immediately.