Push ad creative testing: how to compare creatives by conversions, not clicks

12 min readUpdated Nov 16, 2026

Push ad creative testing: how to compare creatives by conversions, not clicks

Push ad creative testing fails for one boring reason: most advertisers pick a winner from the click report. A push notification with a 1.8% CTR looks like a champion until you sort by conversions and find it has spent your budget producing nothing. Clicks tell you the creative got attention. Conversions tell you it got the right people to act. Those are different questions, and only one of them decides whether you make money. This guide covers how to structure push ad creative testing so every creative is judged on conversion data from your own tracker — not the network's dashboard, not the CTR column, not your gut after two hours of traffic.

Why do push creatives win on clicks and lose on conversions?

Because a push notification's click is a proxy for curiosity, not intent. The user sees a headline and an icon on a lock screen or in a browser notification stack, taps out of reflex, and lands on your page half-interested. A creative that over-promises — "Your phone is infected", "You won a $500 gift card" — will always win the click race. Whether it converts depends entirely on how well your landing page absorbs that curiosity and whether the offer matches the emotional hook.

This creates three specific traps in push ad creative testing:

  1. The curiosity gap is unbounded. Push copy has room for maybe 40–60 characters. It is very easy to write something so vague or alarming that it earns clicks from people who will never convert, and very hard to tell from CTR alone that this is happening.
  2. Networks optimise toward clicks by default. Most push networks will happily rotate your creatives to maximise CTR or spend unless you feed conversions back. If your postback isn't wired, the network's algorithm is learning the wrong objective.
  3. Small samples lie loudly. A creative with 300 clicks and 2 conversions has a conversion rate you should not trust. Push CTRs are high enough that click volume accumulates fast, which tempts you to call winners long before conversion data is meaningful.

The fix isn't complicated. It's discipline: separate tracking per creative, wait for the right signal, and cut on conversions per unit of spend.

Set up per-creative tracking before you launch anything

You cannot compare creatives by conversions if you can't attribute a conversion to a specific creative. On push networks, creative identity is carried by a macro — usually a creative ID token the network substitutes into your landing URL. Your job is to map that token into a sub-ID your tracker records, then read conversions grouped by that sub-ID.

The general pattern, which applies to PropellerAds, RichAds, EvaDav and most others:

  1. Find the creative macro in the network's docs. It's typically something like {creative_id}, {banner_id} or {ad_id}. The exact token differs by network — see Ad network tracking macros explained: what tokens are and how to map them to sub-IDs for the mapping table.
  2. Create one tracking link per creative in Adtraxo. Don't rely on a single link with a dynamic parameter if you can avoid it — a dedicated link per creative gives you a clean, named row in your reports and removes any chance of a macro failing silently and dumping traffic into an "unknown" bucket.
  3. Pass the creative token into a sub-ID. Your landing URL becomes something like https://yourdomain.com/lp?sub1={creative_id}, with Adtraxo capturing sub1 and storing it against the click. Sub-IDs sub1–sub5 are a Pro feature — on the Free plan you get basic stats but not the sub-ID layer, which is exactly the layer this workflow depends on.
  4. Set up the conversion postback. Your offer or affiliate network fires an S2S postback to Adtraxo on conversion, and Adtraxo records it against the click — and therefore against the creative. Postback conversion tracking is also Pro.
  5. Name creatives consistently. Use a scheme like push_ge_2501_angleA_img3 so that when you're staring at a report of 40 rows you know which is which without opening the network.

If you're on PropellerAds specifically, the zone and creative tokens both need to travel — the PropellerAds zone blacklist guide covers that side of it. For RichAds, the RichAds tracking setup guide walks through postbacks and creative IDs for push campaigns end to end. And if you're running native alongside push, the same per-creative logic applies to teasers — the MGID tracking setup guide shows how widget and teaser IDs map into sub-IDs.

What metric should you judge a push creative on?

Judge on conversions per unit of spend — usually CPA or, if your payouts vary, revenue per click — with a minimum sample gate. Not CTR, not conversion rate alone, and definitely not clicks.

Here's why CPA rather than conversion rate. Two creatives can have identical conversion rates and wildly different economics if one attracts clicks from cheap inventory and the other from expensive inventory. CPA folds in both the conversion rate and the cost of getting that traffic, which is the only number that touches your bank account. Conversion rate is a useful diagnostic — if it collapses, the creative-to-offer match is broken — but it isn't the decision metric.

A practical decision framework:

  • Minimum sample before judging: don't call a creative until it has accumulated enough clicks that a single conversion wouldn't flip your conclusion. In practice, for typical push conversion rates, that means hundreds of clicks per creative, not dozens. If you don't know your baseline conversion rate yet, run a small calibration batch first and let the data tell you.
  • Cut rule: if a creative's spend has passed roughly 1.5–2× your target CPA with zero or one conversion, cut it. You don't need statistical certainty to stop spending on a loser — you need a spending limit.
  • Scale rule: a creative that beats target CPA over a meaningful sample gets more budget, not a victory lap. Push performance decays as you scale into colder inventory, so re-check after each budget increase.
  • Hold rule: creatives in between get left alone. The most common self-inflicted wound in push ad creative testing is cutting a creative at 200 clicks because it "looks bad" and never learning whether it would have recovered.

One more thing: check your click counts against the network's before you trust any of this. If your tracker and the network disagree by a wide margin, your conversion attribution is built on sand. Why your ad network and your tracker show different click counts explains the usual causes.

How to structure a push ad creative test so the data is readable

Test one variable at a time, and give each creative its own tracking link. The structure that produces clean data:

  1. Fix the offer and the landing page. If you change the LP mid-test, every creative's numbers become incomparable. Lock the funnel.
  2. Fix the audience and bid. Same geo, same device targeting, same bid, same zones if possible. If the network rotates creatives across different inventory pools, you're comparing audiences, not creatives.
  3. Vary one creative element per round. Round one: headline angle only, same image, same icon. Round two: winning headline, vary the image. Round three: winning headline and image, vary the icon or the CTA. This is what makes the results actionable — you learn which element moved conversions, not just that creative #7 won.
  4. Run 4–8 creatives per round. Fewer and you're not exploring enough; more and your budget fragments across too many small samples.
  5. Let each creative accumulate its own conversion data. This is where Adtraxo's per-campaign analytics with geo and device breakdown earns its keep — you can see whether a creative's conversions are concentrated in one geo or one device type, which often explains a split you'd otherwise misread as creative quality.

If you want a faster read on which clicks actually loaded the page in a real browser, add the one-line arrival snippet to your landing page — it's available on any Adtraxo plan and separates "clicked" from "actually arrived", which is a frequent source of phantom CTR in push.

Feed conversions back to the network, or the network optimises for the wrong thing

This is the step most advertisers skip, and it quietly ruins push ad creative testing. If the network only sees clicks, its auto-optimisation will favour high-CTR creatives regardless of whether they convert. You end up with a campaign that looks healthy in the network dashboard and loses money in your tracker.

The fix is to send conversions back via postback so the network's algorithm optimises toward your actual objective. The mechanics differ per network:

  • PropellerAds: postback with the click ID and, where supported, the zone and creative tokens so the network can learn at the creative level.
  • RichAds: see the RichAds setup guide for the exact postback URL structure and creative ID handling.
  • Taboola: if you're running native alongside push, the Taboola conversion tracking guide covers the postback flow for affiliates.

While you're in the network, keep an eye on placement-level waste. A creative can look mediocre because it's being shown on terrible inventory, not because the creative is bad. The Taboola publisher blocklist guide and the MGID widget blacklist guide cover how to prune placement waste before you blame the creative.

Use fraud and arrival data to stop bad creatives polluting your results

Push traffic attracts bots, and bot clicks inflate CTR without ever converting. If one creative is being served disproportionately to bot-heavy inventory, it will look like a click winner and a conversion loser — and you'll wrongly conclude the creative is bad when the traffic was.

Adtraxo's fraud detection — IP velocity, datacenter IP, bot user agent, uniqueness conflict and invalid referer checks — flags suspicious clicks so you can exclude them from your creative comparison. On Pro, you can also set fraud rules to block them outright. The practical workflow: before you cut a creative, check whether its clicks are disproportionately flagged. If they are, the creative hasn't been fairly tested.

Combine that with the arrival snippet: a creative with a high click count but a low arrival rate is being clicked by something that isn't loading your page. That's a traffic-quality problem, not a creative problem, and cutting the creative won't fix it.

When should you stop testing and lock in a winner?

When a creative has beaten your target CPA over a sample you trust, across at least two budget levels, and its conversions aren't concentrated in a single fragile geo or device. At that point, promote it to your control and start a new round varying a different element.

Keep the losing creatives in a folder, not the bin. Angles that failed on one offer frequently win on another, and push creative fatigue is real — the winning creative will decay, and you'll want a bench. A simple rotation: every time the control's CPA drifts above target for a sustained stretch, swap in a challenger from the bench and re-run the comparison.

And revisit your winners when you change networks. A creative that wins on PropellerAds inventory may not win on EvaDav, because the audience and the notification context differ. The per-creative sub-ID structure makes that comparison trivial — same link schema, different network macros.

If you're setting up tracking across several push and native networks at once, the network-by-network tracking setup guide is the right starting point before you build your first creative test.

Frequently asked questions

How many clicks does a push creative need before I can judge it by conversions?

There's no universal number, because it depends on your offer's conversion rate. The rule is to wait until one additional conversion wouldn't flip your conclusion. In practice that means hundreds of clicks per creative for typical push conversion rates. Until then, apply a spend-based cut rule (roughly 1.5–2× target CPA with zero or one conversion) rather than declaring a winner.

Should I optimise push creatives on CTR at all?

Only as a diagnostic. CTR tells you the creative earns attention; it says nothing about whether the traffic converts. Use CTR to spot creatives that are being ignored, but make keep/cut/scale decisions on conversions per unit of spend. A high-CTR, zero-conversion creative is the most expensive kind.

Why does my tracker show fewer conversions than the ad network?

Usually because the postback isn't firing correctly, the click ID isn't being passed through, or the network is counting a different event (such as a landing page view) as a conversion. Check that your postback URL contains the correct click ID macro and that conversions appear in your tracker's report against the right sub-ID. Click discrepancies are covered separately in our guide on tracker vs network click counts.

Can I run push ad creative testing on the Adtraxo Free plan?

You can track clicks and see basic stats on Free, but the workflow in this guide depends on sub-IDs (sub1–sub5) and S2S postback conversions, which are Pro features. If you want per-creative conversion attribution, you need Pro at $27/month (founding rate; it rises to $47 for new accounts, and existing subscribers keep $27). It includes a 30-day money-back guarantee and you can cancel any time.

How often should I replace a winning push creative?

When its CPA drifts above target for a sustained stretch, not on a fixed calendar. Push creative fatigue shows up as slowly rising CPA and falling conversion rate on the same creative and audience. Keep a bench of tested challengers so you can swap without pausing the campaign.

Ready to stop guessing which push creative actually pays? Set up per-creative tracking links, wire your postbacks, and let conversions pick the winner. Start at adtraxo.com.

Track your affiliate campaigns for free

Clicks, conversions, fraud detection — all in one tracker.

Get started free →

Not ready for an account? Get the Tracking Domination Playbook — free, just your email.