When a winning playable starts to fatigue, don't replace it. Change one thing at a time, starting with the parts players see first: the hook, then the tutorial hint, then difficulty, then the win or fail moment, the end card, and the skin. Leading studios spend about 70% of their playable output on variations of what's already winning and about 30% on new concepts. On the fastest networks, a playable often gets one to two weeks before performance slides, so the next version should be ready before you need it.
- Fatigue is normal. Practitioners report seven to ten good days on the fastest channels and one to two weeks on Meta.
- Iterate the winner before you invent a new concept. Only about 5% of creatives become real winners.
- Change one variable per version so you know what moved the number.
- Work from the outside in: hook, tutorial, difficulty, win or fail, end card, skin, then a new mechanic.
- With Genie Playable, your team makes the next version from the last one. The next version, not a new brief.
How long does a playable ad last before it fatigues?
There's no official number, and no network publishes one. What practitioners report is consistent enough to plan around:
| Channel | What practitioners report | Source |
|---|---|---|
| Fastest in-app networks | About seven to ten solid days before performance slides | AppSamurai |
| Meta | Fatigue within about one to two weeks | AppAgent |
| Meta and TikTok, high spend | Decline can start in three to seven days; rotate every seven to ten | Segwise |
A UA lead at a large studio described the other side of this to us: they now produce more playables than UA can absorb, and the goal has moved from quantity to better metrics. Volume alone doesn't fix fatigue. The right next version does.
Should I make a new playable or iterate the winner?
Iterate first. Winners are rare, and a proven concept already carries most of what made it work. Sett's analysis of top studios shows the pattern clearly:
- At one studio, 70% to 75% of playables were deliberate variations of a winning concept. The other 25% to 30% explored new directions.
- Whiteout Survival ran 42 playables on AppLovin in October 2025: 15 concepts and 27 variations. The winning concept alone got 11 variations.
- The top 20 grossing games shipped a median of 64 new playables a month.
The real question isn't "new or old". It's which single change will buy the winner another two weeks.
What should I change first when a playable fatigues?
Test one variable at a time. Change the hook and the end card together, and you won't know which one worked. Start with what players see first, because that's where most of them leave. Sett's data from 3.5 million sessions shows median drop-off of about 47% in the first three seconds when there's no guided interaction.
| Step | What to change | What it tests | Signal to watch |
|---|---|---|---|
| 1. Hook | The first two to three seconds: lead-in video, opening frame, the problem on screen | Whether players start at all | Engagement and first-three-second drop-off |
| 2. Tutorial hint | The hand, arrow or prompt that shows the first move | Whether players understand what to do | Time to first interaction |
| 3. Difficulty | How hard the challenge is, how fast it resolves | Whether players keep going | Completion and mid-game drop-off |
| 4. Win or fail | Let them win, let them nearly win, or stop mid-game | Which ending drives the install | CTA click rate after win vs after fail |
| 5. End card | Copy, button, character, offer | Whether interest turns into a click | End card to CTA rate |
| 6. Skin | Character, theme, current live event | Whether a fresh look resets fatigue | IPM against the previous version |
| 7. Mechanic | A different core interaction | This is a new concept, not a variation | Treat it as part of your 30% exploration budget |
"Looks clever and almost always tests worse."
If you want to try a second mechanic, ship it as its own version.
How do I know a variation actually worked?
Give it enough data. A common rule of thumb is about 50 conversions per creative in four days, or 100 in seven, before you call it. On AppLovin, don't pause creatives by hand too early; the algorithm shifts spend on its own.
Judge on the metric that pays the bills. Senior UA people are blunt about this: CPI on its own is misleading, and D7 ROAS and retention are what matter. In one public test of 360 creatives, the playable with the best IPM wasn't the one that paid back best. Use in-playable signals (engagement, completion, time to first tap) to decide what to change next, and ROAS to decide what stays live.
How do I keep a steady flow of variations?
- Keep a variation backlog for every live winner: three hooks, two difficulty levels, two end cards, the next event skin.
- Tie skins to your LiveOps calendar. A playable wearing this week's event looks fresh and matches what new players find in the game.
- Set a cadence. AppAgent suggests three to six new playables a month for Meta. AppSamurai suggests two to three concepts plus six to ten variants a week per title at higher spend.
- Log every change with its result, so the next person doesn't retest what already failed.
"The idea is the easiest part. We have too many ideas."
The bottleneck is rarely ideas. It's how long each change takes to become a playable you can upload.
Where does Genie fit?
Genie Playable is built around the next version. Your team builds a playable from your game (your APK or HTML build plus the assets, narrative and live events in Genie), from a format that's already working, or from scratch from a brief.
When it fatigues, you change one thing on the last version and export it again for Meta, Google, TikTok, AppLovin, Unity Ads, Mintegral, Liftoff, Moloco, Snapchat and Digital Turbine. Genie Core keeps every variation on brand, so version twelve still looks like your game. Try it free on your game.
FAQ
How long does a playable ad last?+
Practitioners report about seven to ten good days on the fastest in-app networks and one to two weeks on Meta. No network publishes an official number.
How many variations should I make of a winning playable?+
As many as keep working. Leading studios put about 70% of playable output into variations of winners. Whiteout Survival gave one winning concept 11 variations in a single month.
What should I change first in a fatiguing playable?+
The hook. The first two to three seconds decide whether players start at all. Then the tutorial hint, difficulty, win or fail moment, end card and skin.
Should I test more than one change at a time?+
No. Change one variable per version so you know what caused the result.
Which metric should I use to judge a playable variation?+
D7 ROAS and retention for keeping it live. In-playable metrics like engagement and completion for deciding what to change next. CPI on its own is misleading.
Sources
- Sett, Playable Ads Complete Guide
- Sett, State of Playables Q4-25
- AppSamurai, playable ads creative testing
- AppAgent, what are playable ads
- AppAgent, State of Playable Ads 2025
- Segwise, creative fatigue
- Playio, creative fatigue in mobile game ads
- Matej Lancaric, finding a winning creative
- RocketshipHQ, testing creatives on AppLovin