How to test ad creatives on AppLovin in 2026: upload 5 to 7 new videos per week as creative sets inside your existing campaigns, keep 2 to 3 active concepts at any time, refresh every 2 to 3 weeks, and let Axon 2 allocate budget across them.
Do not run a separate test campaign. Do not pause underperformers manually (Axon deprioritizes them for you). The testing mental model is fundamentally different from Meta and TikTok, where teams ship 50 to 100 variants per week and prune aggressively.
In 15 years of running mobile UA, with $100mm+ deployed across 100+ apps including significant AppLovin spend, the most expensive pattern I see on AppLovin is teams applying their Meta playbook. They ship 30 to 50 variants a week. They prune aggressively. They restructure campaigns every few days.
None of that works on AppLovin. Axon is a fundamentally different algorithm. The teams that win on AppLovin run fewer, more concentrated tests and resist the urge to touch the campaign.
Page Contents
How do you test ad creatives on AppLovin in 2026?
Test ad creatives on AppLovin by uploading 5 to 7 new videos per week as creative sets inside your existing campaigns, keeping 2 to 3 active concepts live at any time, refreshing every 2 to 3 weeks, and letting Axon 2 allocate budget across them. There is no separate test campaign. There is no manual pruning. The mental model is hands-off iteration on a low-variant, concentrated-spend basis. Axon needs concentrated spend per creative to calibrate; spreading thin breaks the learning loop.
Three structural elements define a working AppLovin testing process:
- Low variant volume. 5 to 7 new creatives per week, not 50 to 100. AppLovin’s own platform guidance matches this. At high-volume scaled accounts, the ceiling stretches to roughly 100 per month (about 25 per week), but the baseline cadence is 5 to 7.
- Concentrated spend per creative. Each active creative needs enough daily budget for Axon to read its signal. Spreading $5K/day across 30 variants is worse than concentrating it on 3 to 5. The algorithm cannot calibrate at low per-variant spend.
- Hands-off iteration. Do not pause underperformers, do not change campaign goals, do not restructure budgets every two days. Axon’s learning curve assumes stability. The teams that get this right resist the impulse to optimize.
For the testing mental model on Meta and TikTok (which is the opposite of this), see how to test ad creatives on Meta in 2026 and how to test ad creatives on TikTok in 2026.
How do you add a new creative to test on AppLovin?
Add a new creative to test on AppLovin by uploading it as a new creative set inside an existing campaign, or by cloning an existing creative set and adding the new asset. The clean structure is one creative set per concept, with each set containing the variants of that concept (different first frames, end cards, copy overlays). New creatives slot into the existing campaign. You do not isolate them in a separate test campaign as you would on Meta.
The 80/20 rule AppLovin’s own documentation recommends:
- One clear objective per campaign. Do not mix install volume goals with ROAS goals in the same campaign.
- One main audience cluster per ad set. Mixing audience targets dilutes Axon’s signal.
- Multiple creative concepts in each creative set. Give Axon room to find winners without creating chaos in the account structure.
For early visual evaluation before you upload to AppLovin Ads Manager, RocketShip HQ’s free AppLovin ad preview tool gives you a drag-and-drop mockup of interstitial and rewarded formats. Useful for evaluating end-card legibility and first-frame composition before the creative goes live.
What is Axon 2 and how does it pick winners?
Axon 2 is AppLovin’s AI advertising engine. It processes more than 2 million ad auctions per second across over 1 billion devices in 2026. It picks creative winners through a continuous learning loop trained on user-level behavioral signals, particularly ad monetization bid data from MAX-integrated apps. Unlike Meta’s Bayesian-style budget allocation across many variants, Axon concentrates spend on a small number of creatives and learns deeply on each. The algorithm auto-deprioritizes losers without manual intervention.
What Axon does differently from Meta and TikTok:
- Learns from ad monetization signal, not just install conversion. Through MAX, Axon sees the bid prices users command on the ad-mon side. This becomes a predictor of which users monetize well downstream.
- Concentrates spend rather than spreading it. Axon’s prediction accuracy increases with spend volume per creative. Meta’s algorithm is more forgiving on per-variant spend.
- Continuous learning, not batched. Updates happen in real time as user behavior accumulates, not in periodic optimization cycles.
The practical implication: trust Axon. The instinct to manually rebalance budgets, pause underperforming creatives, or restructure campaigns every few days actively hurts performance on AppLovin. The algorithm is built for stability.
Should you optimize for D0, D7, or D28 ROAS on AppLovin?
Choice between D0, D7, and D28 ROAS optimization on AppLovin depends on your app’s monetization model. D0 ROAS works for IAA-heavy gaming apps where revenue lands fast through ad views. D7 ROAS works for hybrid monetization (ad views plus light in-app purchases). D28 ROAS works for IAP-heavy games and subscription apps with longer payback windows. The wrong choice forces Axon to optimize against signal that does not yet exist for your app, which leads to over-spending on the wrong users.
| Monetization model | Recommended ROAS window | Why |
|---|---|---|
| IAA-heavy gaming (ad-monetized) | D0 | Revenue from ad views accrues immediately; Axon can read signal on day one |
| Hybrid (IAA + light IAP) | D7 | Balances early ad revenue with first-week IAP signal |
| IAP-heavy gaming | D28 | Most IAP revenue lands in the first 28 days; D0 / D7 windows would mis-target |
| Subscription app (non-gaming) | D28+ | Trial-to-paid windows mean most revenue lands later; the signal needs time to mature |
Setting the wrong ROAS window is one of the most common AppLovin mistakes, surfaced repeatedly by guests on the Two and a Half Gamers podcast episode on scaling AppLovin UA with ex-AppLovin advisor Alexis Lejeune.
How is the creative grammar different on AppLovin?
Creative grammar on AppLovin is different because most AppLovin ad inventory is not dismissible. Rewarded video plays to completion because the user opted in for an in-app reward (virtual currency, extra life, premium content). Interstitials run for roughly 5 seconds before the close button appears. Unlike Meta and TikTok, where the user can scroll past in the first 1 to 3 seconds, AppLovin gives you the full ad duration with the viewer’s attention captured. The hook is less critical; immersive storytelling, gameplay reveals, and reward visualization matter more.
On Meta and TikTok, the first second of the creative decides most of the variant’s performance because the user can scroll past at any moment. The 3-second rule on TikTok and the 2 to 3 second hook window on Meta are operational responses to dismissible inventory. If the hook does not land immediately, the user is gone.
AppLovin’s dominant formats invert that dynamic. Rewarded video is the highest-revenue UA format on the platform, and by design the user has chosen to watch it in exchange for an in-app reward. They are motivated to complete the view. Interstitials run with a delayed close button, so the first few seconds are protected attention. Playable ads turn the ad time into engagement time, with the user actively interacting rather than passively watching.
The creative implications are structural:
- Hook tactics that win on Meta and TikTok matter less. Cold opens, curiosity gaps, and pattern interrupts in second one are not the load-bearing decision on AppLovin.
- The middle and end of the ad carry more weight. Users will see them, so build the strongest payoff there, not in the first second.
- Gameplay reveals and reward visualizations belong further into the ad. The format gives you 15 to 30 seconds of captured attention; use the full duration.
- Pacing can be slower. You do not need to compress everything into the opening seconds. Build the story.
- Playables maximize this captured attention. The interactive format converts the watch time into engagement time, which Axon reads as qualification signal.
This does not mean the hook is irrelevant on AppLovin. It means the hook is one of several load-bearing elements, not the single load-bearing element. Spend creative budget on full-ad immersiveness rather than concentrating it on the opening second.
Why do playable ads matter more on AppLovin?
Playable ads matter more on AppLovin because they serve a user-qualification function the algorithm uses for downstream prediction. A user who interacts with a playable ad has signaled both intent and ability to engage with the app’s core mechanic before installing. That interaction is signal Axon uses to predict LTV. On Meta or TikTok, playables are an option among many; on AppLovin, they often are the highest-converting format for gaming.
What makes playables work on AppLovin specifically:
- Pre-install qualification. The user has demonstrated they understand and engage with the gameplay before they install. Axon weights this signal heavily.
- Higher CPM tolerance. Playable inventory carries higher CPMs because publishers earn more per impression on engaged formats. Axon’s MAX visibility makes this tradeoff economical.
- Lower install-to-action drop-off. Users who installed after a playable are more likely to complete the in-app event you optimize for, because they have already simulated the experience.
Playables are not optional for serious AppLovin spend on gaming apps. They are the format that does the heaviest lifting on user qualification, and the platform’s economics reward them.
What is a good creative refresh cadence on AppLovin?
A good creative refresh cadence on AppLovin is every 2 to 3 weeks per winning concept. That is slower than Meta or TikTok refresh cycles (10 to 14 days on TikTok, 14 to 21 days on Meta), which is counterintuitive because the platforms feel similar from the outside. The reason: Axon’s concentrated-spend learning model means a single winning creative can sustain longer before audience saturation hits, because the algorithm is precise about which users see it.
Signals that a winning creative is fatiguing on AppLovin:
- Frequency above 4 per unique user in the last 7 days
- CPI rising 15% or more on a 7-day rolling window without a corresponding LTV signal
- D7 retention on the cohort acquired by that creative dropping 10% or more
- Creative CPM staying flat while account CPM rises (Axon is paying more elsewhere)
The 5 to 10 creatives per week guidance at $10K/day spend, surfaced by Alexis Lejeune on the Two and a Half Gamers episode, is the operational anchor for refresh velocity at scaled accounts. At lower spend, scale down proportionally.
What budget do you need to test creatives on AppLovin?
Budget on AppLovin scales by your cost-per-purchase target. The working formula is daily budget = D7 cost per purchase × 10 to 15, so you generate 10 to 15 daily conversions for Axon to calibrate. For an app with $150 D7 cost per purchase, that is roughly $1,500 per day per campaign. For an app with $800 D7 cost per purchase, that is $8,000 per day. $500 per day is the absolute floor; below it, calibration is inconsistent. D7 ROAS campaigns calibrate in 3 to 5 days; D28 ROAS campaigns need 10 to 14 days.
The formula-driven version, per AppLovin’s own guidance and Alexis Lejeune on the Two and a Half Gamers episode:
| Your D7 cost per purchase | Minimum daily budget per campaign | Why |
|---|---|---|
| $50 | $500 – $750 | Absolute floor; targets 10-15 conversions/day |
| $150 | $1,500 – $2,250 | 10-15 daily conversions for Axon to learn |
| $400 | $4,000 – $6,000 | Higher CPPD7 requires proportionally higher spend |
| $800 | $8,000 – $12,000 | IAP-heavy gaming; D28 calibration takes 10-14 days |
| $1,500+ | $15,000+ | Subscription apps and high-LTV gaming; longer payback windows |
The target is 50 to 100 installs or conversions in week one of a new campaign. Below that, Axon does not have enough signal to make meaningful predictions. Spending below the formula floor is the most common reason AppLovin “does not work” for an advertiser; the algorithm cannot calibrate at low per-campaign spend.
To back-calculate the testing budget from your spend, CPA, and refresh-rate targets, use RocketShip HQ’s free creative testing calculator. For broader ROAS modeling across paid spend, see the ad cost calculator.
Why does AppLovin work better for gaming than non-gaming?
AppLovin works better for gaming than non-gaming because the MAX SDK is widely integrated in gaming apps but rarely in non-gaming apps. MAX gives Axon visibility into ad monetization bid levels in real time, which is the strongest signal Axon uses to predict which users will monetize well downstream. Without MAX, Axon’s predictions degrade. Small and medium non-gaming advertisers often cannot launch campaigns at all on AppLovin without a critical mass of spending and conversion data first.
The AppsFlyer Performance Index 2025 ranks AppLovin clearly on the gaming side:
- #1 in iOS gaming, Tier 1 North America and Western Europe. AppLovin overtook Apple Ads in this tier as the leading UA source for iOS gaming.
- #1 in the inaugural Creative Performance Index for gaming. Driven by first-place finishes in gameplay-led creative and creatives combining animated and real-life footage.
- Top 3 in Android gaming globally. Behind Google Ads on volume but gaining share year over year.
- Outside gaming. AppLovin is not ranked in the top tier for non-gaming categories, with the structural limitation traced to MAX absence and audience-fit mismatch.
See the full AppsFlyer Performance Index 2025 for the underlying methodology and per-category rankings. The Creative Performance Index addition in 2025 is the first time AppsFlyer separately ranked networks on creative performance, and AppLovin’s #1 ranking there reflects the gameplay-led, full-ad-duration creative grammar covered above.
For subscription apps and other non-gaming categories, AppLovin should be one channel in a diversified mix, not a primary scale lever. The broader subscription growth framework is covered in how to grow a subscription app.
How does the MAX SDK integration affect creative performance?
MAX SDK integration affects creative performance on AppLovin because it gives Axon visibility into ad monetization bid levels in real time. Apps with MAX integrated see dramatically better UA performance because Axon can predict which users will monetize at high CPMs downstream. The integration creates a flywheel: more MAX adoption means more bid signal for Axon, which means stronger UA performance, which means more advertisers want to integrate MAX. This is the structural moat behind AppLovin’s gaming dominance.
What this means for your creative testing:
- If you have MAX integrated. Axon’s creative-winner picks will be biased toward users who command higher ad-mon bids. The variants that “win” are not just the variants with the best CPI; they are the variants that bring in the most monetizable users.
- If you do not have MAX integrated. Axon falls back to thinner signal (mostly install conversion plus reported events). Performance will be measurably lower, and your creative testing will produce noisier reads.
- If you are on a non-gaming app where MAX is uncommon in your category. Axon cannot rely on the same signal stream. Expect higher CPIs, lower predictability, and more variance in creative testing reads.
Integrating MAX is not technically a creative decision, but it is the load-bearing variable behind your creative testing economics on AppLovin.
What are common AppLovin creative testing mistakes?
Five common mistakes that show up repeatedly across gaming and subscription app accounts on AppLovin:
- Applying the Meta playbook. Shipping 30 to 50 variants a week breaks Axon’s calibration. The algorithm cannot learn from low per-variant spend. Stay at 5 to 7 new creatives per week, scaling to ~25/week only at very high spend tiers.
- Pausing underperformers manually. Axon deprioritizes losing variants automatically. Manual pausing interrupts the learning signal and forces Axon to relearn. Trust the algorithm.
- Touching campaign goals every 2 days. Goal changes reset Axon’s learning. Set the goal per geo, leave it alone, and let the algorithm converge. The patience is uncomfortable but required.
- Running on $500 daily budgets. Axon cannot calibrate at sub-$1,000 per day for IAA games, or below $5,000 for IAP games. Below the budget floor, your “tests” are noise.
- Ignoring playables. The playable is the format that does user qualification for Axon. Apps that only run video on AppLovin leave performance on the table. Build playables into the creative production pipeline.
For the deeper operational context on the gaming-specific tactics, see the AppLovin playbook. For the broader cross-platform creative testing framework, see the A/B testing framework for ad creative at scale.
Frequently asked questions
What is the difference between AppLovin and Axon?
AppLovin is the company; Axon is the AI advertising engine AppLovin built. Axon powers AppDiscovery (AppLovin’s UA platform for advertisers) and increasingly other AppLovin products. When operators say “Axon” they usually mean the algorithm and prediction model. When they say “AppLovin” they usually mean the broader platform including AppDiscovery, MAX, and the rest of the stack.
Do you need MAX SDK integrated to advertise on AppLovin?
No, you do not need MAX SDK integrated to advertise on AppLovin, but you should expect significantly weaker performance without it. MAX gives Axon visibility into ad monetization bid levels in real time, which is the strongest single signal Axon uses to predict downstream user value. Apps with MAX integrated consistently outperform apps without, often dramatically.
How long does Axon take to learn for a new creative?
Axon typically takes 3 to 7 days to calibrate a new creative at the right spend level. Below the spend floor (roughly $1,000/day for IAA games, $5,000/day for IAP games), calibration takes longer or never completes. The algorithm needs concentrated impressions per creative to read user-value signal; thin spend slows learning.
Can you A/B test creatives on AppLovin?
AppLovin does not have a Meta-style A/B test tool with audience-level holdouts. The way you “test” on AppLovin is by uploading new creatives as creative sets in your existing campaign and letting Axon allocate budget across them. Maintain 2 to 3 active concepts at any time so Axon has variant choice without dilution. There is no built-in split test feature; the algorithm handles variant comparison through budget allocation.
What hooks work best on AppLovin?
For gaming, hooks that preview core gameplay loops or the most rewarding moment work best (level wins, satisfying mechanics, character progression). For non-gaming on AppLovin, outcome-led hooks similar to TikTok and Meta work, but with the caveat that AppLovin’s audience leans heavily toward gaming users, so non-gaming creative needs to bridge the audience-fit gap. Playables outperform pure video for most gaming categories.
How does AppLovin’s e-commerce pilot work for non-gaming brands?
AppLovin’s e-commerce pilot, launched in 2025, extends Axon’s machine learning to non-gaming advertisers (retail, DTC, subscription). It uses behavioral signals from AppLovin’s gaming-heavy audience to find purchase-likely users. Early results have been mixed: some brands see strong performance, others find the inventory mismatched to their audience. The pilot is worth testing if you have meaningful budget but should not be a primary channel until your category has demonstrated fit.
Should you use AppLovin alongside Meta and TikTok or instead of them?
Use AppLovin alongside Meta and TikTok, not instead of them. AppLovin is strongest on gaming and serves a specific algorithmic niche (concentrated spend, ad-mon signal, playables). Meta and TikTok serve broader audiences with different creative grammars. The right channel mix depends on your vertical and spend tier; for gaming, AppLovin can be the largest single channel, but it works best when paired with Meta and TikTok for diversification.
What ad formats does AppLovin support?
AppLovin supports three primary ad formats: video (interstitial and rewarded), playable ads (interactive demos), and interactive ads (a hybrid between video and playable). Banner formats exist but are not typically used for UA. Most successful AppLovin advertisers run a mix of video and playable, with playables doing the bulk of user qualification on gaming campaigns.
How much does it cost to advertise on AppLovin?
AppLovin CPIs in 2026 typically range from $1 to $8 for gaming apps depending on category, geo, and competition, and $3 to $20+ for non-gaming apps where MAX integration is rare. The minimum viable daily budget per campaign is around $1,000 for IAA gaming and $5,000+ for IAP or subscription apps. Account-level minimums for full Axon access also apply for non-gaming advertisers.
