App install ad creative lives or dies in the first two seconds, before a logo, a name or a brand promise has registered with anyone. What belongs in that window is the app doing the one thing the viewer wants: the number changing, the match appearing, the transfer completing, the mess becoming organised. Everything else, including your name, goes after the viewer has already decided they want the outcome, and the fastest way to lift install rate is usually to cut the first two seconds you currently have rather than to add anything.
What goes in the first two seconds?
Pick the app's payoff, not its entry point. Most install ads open on a login screen, an empty state or an onboarding carousel, which is the app at its least appealing and the exact moment a real user is most likely to abandon it.
| App category | Weak open (common) | Strong open (2 seconds) |
|---|---|---|
| Fintech and banking | Logo, then a sign-up screen | A balance or a saving figure changing on screen |
| Budgeting | Dashboard overview | The subscription the app just found and cancelled |
| Health and fitness | Person in a gym | A tracked graph crossing a threshold |
| Language learning | App icon and tagline | Someone answering correctly, with the streak counter ticking |
| Dating | Brand promise text | A match landing, then the first message |
| Productivity | Empty project board | A chaotic list resolving into an ordered one |
| Photo and video tools | Before shot alone | The before-to-after transition, mid-swipe |
| Delivery and marketplace | Category grid | The item found, then the order confirmed |
Two ideas run through the right-hand column. First, motion in the interface beats a static screen, because a change on screen is a promise about the future. Second, the payoff should be small and concrete. "Cancel the subscription you forgot about" installs better than "take control of your finances", because the viewer can picture it.
Sound is optional and captions are not. Assume the first viewing is silent, so the claim has to be readable in the frame.
How do you make screen capture legible in a feed?
This is the production skill that separates app creative from every other kind, and most teams get it wrong by trying to be faithful to the product.
The problem is scale. Your interface was designed for a phone at arm's length. In a feed it is rendered inside a phone, on a phone, sometimes at a quarter of the screen height. A 15 point label is unreadable. A full screen recording shows nothing.
What works:
- Crop to one element. One card, one counter, one row, filling 60 to 80 percent of the frame. Show the whole screen only if the whole screen is the point.
- Scale the UI up before recording. Use a staging build with a larger base font size, or the device's own accessibility text size, rather than zooming a finished recording and getting soft edges.
- Capture at 60 frames per second and slow motion-heavy sequences slightly in the edit. Animation that feels quick in hand reads as a flicker in a feed.
- Populate a realistic demo account. Zero balances, placeholder names and empty lists make an app look unused. Fill it with plausible data, and redact anything real.
- Show the tap. A finger, a subtle press state or a highlight ring. Without it, the interface appears to change by itself and the viewer cannot see cause and effect.
- Drop device frames in vertical placements. A phone bezel inside a 9:16 ad costs you a third of the pixels you paid for.
- Clean the status bar. Full battery, no personal notifications, no debug banners.
Legibility is measurable, not aesthetic. View the creative at 25 percent size on a laptop. If you cannot tell what changed, no one scrolling can either.
Why should you vary app moments rather than captions?
Because a caption change does not change what the viewer is deciding. A set of twelve files with twelve hook lines over the same screen recording is one argument in twelve costumes, and it fatigues on one curve.
Our 2026 fatigue benchmark, a synthesis of published platform and agency figures rather than our own measurement, puts the CTR decline at 15 to 20 percent in a creative's first two weeks, with week three landing 45 to 70 percent below launch and week five averaging 38 percent below peak. Meta internal research in the same report shows a 45 percent CTR drop after a fourth exposure to the same creative. Caption variants do not reset that, because exposure is to the visual.
Real variation means a different moment of the app. An inventory to work from:
- the onboarding moment where the app first becomes useful
- the number that changes (saved, earned, tracked, completed)
- the notification that arrives at the right time
- the streak or history view, which sells persistence
- the search that returns exactly the right result
- the share or export, which sells the output rather than the app
- the recovery moment, where the app fixes a mistake
Seven moments times three or four openings each gives you a set of 8 to 20 live variations, which is what an active campaign typically needs in rotation, without repeating a single visual. Getting that from brief to live in a week is a process problem more than a creative one, and our creative testing workflow page sets out the shape we use.
Hook rate is how you judge the moments against each other. Our hook rate note covers the calculation, and the medians in the benchmark are 33 percent on TikTok, 28 percent on Meta Reels and feed, and 22 percent on YouTube in-stream.
How does pacing differ by network?
Enough to require different cuts, not different captions on one cut.
| Network | Median video CTR | Median hook rate | Frequency ceiling | Pacing that works |
|---|---|---|---|---|
| Meta Reels | 1.78 percent | 28 percent | 2.5 weekly | Fast open, one idea, 8 to 15s |
| Meta feed | 1.62 percent | 28 percent | 2.5 weekly | Slightly slower, captions carry more, 10 to 20s |
| TikTok | 0.84 percent | 33 percent | about 3.0 | Native pacing, spoken voice, cuts every 1 to 2s, 6 to 15s |
| YouTube in-stream | 0.42 percent | 22 percent | 3 to 7 per week | The payoff before the 5 second skip, then a real body, 15 to 30s |
The YouTube column is the one people misread. The lower CTR is a function of the placement rather than the creative being worse, and the skip button makes the first five seconds a hard gate rather than a soft one. Front-load harder there than anywhere else.
Refresh cadence follows frequency, not taste. App install and TikTok sit at weekly, Meta feed tolerates fortnightly, and above a weekly frequency of about 3 you should assume fatigue has already arrived. The report's operational conclusion is that throughput rather than talent is the bottleneck: unique concepts shipped per month predicts campaign longevity better than the quality of any single ad, and brands shipping 15 to 50 variants a month see 3 to 5 times longer campaign lifespan than quarterly refreshers.
Where Genyad fits, and where it does not
Genyad turns footage you already own into publishable ad variations. Upload your screen captures and any filmed footage once, it transcribes and tags every clip, and each variation is a fresh script, shot selection, voiceover, caption set and export drawn from that library rather than a re-cut of one timeline. One standard variation is one credit, editing and re-exporting cost nothing, and exports come in 9:16, 4:5, 1:1 and 16:9 at 1080p with no watermark. Our app install video ads page has the detail.
The gaps are worth stating plainly for UA teams. Genyad does not build playables, does not produce static banners or end cards, has no AI avatars or synthetic presenters, does not predict performance before you spend, and does not publish to any ad network directly. It also cannot record your app for you, so screen capture quality remains your job, and it is the input that most determines the output.
Frequently asked questions
What should an app install ad show first?
The app's payoff in motion: the number changing, the match landing, the task completing. The install decision is made before a logo or a brand line has registered, so the first two seconds should show the outcome the viewer wants rather than the app's entry screen.
How do you make app screen recordings readable in a feed?
Crop to a single interface element filling most of the frame, record at 60 frames per second with an enlarged base font, populate a realistic demo account, and show the tap so cause and effect are visible. Check the result at 25 percent size on a laptop, which is roughly how large it will appear in a feed.
Is it enough to change the hook text between variations?
No. Exposure is to the visual, so caption variants of the same screen recording fatigue on the same curve. Vary the app moment instead: onboarding, the changing number, the notification, the streak, the search result, the export.
How often should app install creative be refreshed?
Weekly at meaningful spend. Weekly frequency of 2.5 on Meta prospecting is where decline begins and around 3 is where fatigue is already present, so app install campaigns generally need new concepts entering rotation every week rather than every month.