The curse of knowing your own product
Launch creative is usually made by people who have thought about the product daily for months. They know the category, the problem and the vocabulary, and it is genuinely difficult to write for someone with none of that context.
The resulting ad answers questions the viewer has not asked yet, and skips the one they actually have, which is what is this.
Establish the category before the product
A cold viewer needs to know what kind of thing this is before they can evaluate whether they want it. Opening on a differentiating feature assumes a frame the viewer does not have — the feature is meaningless until they know what it is a feature of.
Category first, then the specific claim. It feels redundant to the team and it is not redundant to the audience.
Show the interface
For an app, the product is the interface. Abstract lifestyle imagery communicates mood and defers the only question that matters — what will actually happen after the install.
Showing real screens is also a trust signal. Creative that carefully avoids showing the product invites the reasonable suspicion that there is a reason.
One idea per piece
Fifteen seconds accommodates one idea properly or three badly. Launch creative tends toward three, because everyone involved has a favourite feature and the piece becomes a negotiated settlement.
Make several pieces, each committing to one idea. That is cheap now, and it also produces the comparison that tells you which idea was actually the strongest — information a single compromise asset can never give you.
Match the creative to the store listing
The ad is half the journey. A viewer arriving at a listing whose screenshots and first line look unrelated to what they just watched will bounce, and the loss is attributed to the creative rather than to the discontinuity.
Treat the ad and the first screen of the listing as one continuous piece of communication, because to the user that is exactly what they are.