True Take Studio · proof page, not indexed

Meta ad teardown

A US sleep tech brand, September 2026 · public sources only, no access to the account

The brand is not named here. They are not a client, they never asked to be read in public, and what matters on this page is the method, not the company. I have no relationship with them and have never worked on the account. This is here to show how I read an ad account before anyone hands me performance data, because that first read is where a creative brief goes right or wrong.

What I looked at

Meta’s Ad Library, US, the brand’s active ads, pulled 10 September 2026. About 91 ads live, roughly 30 of them video. I read the top of the impression sort, 14 distinct entries, and the brand’s homepage the same day.

What the library proves

All 14 started between 26 August and 3 September 2026. Nothing older sits at the top. That’s a launch push behind a newly released device, not a steady state.

Twelve are brand owned. Two run as branded content through creator handles, so there’s a creator voice in the mix, but it’s a thin slice.

Five primary text lines carry all twelve, each running across two to eight placements. The angles are narrow. Four lead with the device and a general benefit. One opens on the buyer’s state, tired of being tired, and it sits among the most reused lines in the slice.

Where the reading stops

Reuse is where most teardowns quietly cheat. A creative sitting in eight ads has been duplicated across placements, not voted up by results. Reading placement count as performance is the most common error I find in competitor audits, and it sells well because it looks like data.

Spend, click through, cost per acquisition, hold rate, revenue, and the creative actually carrying the account are all invisible here. Anyone who hands you a ranked list of a competitor’s best performing ads off this library is guessing and calling it analysis. With the account’s own numbers I’d settle half of what follows in an afternoon. Without them it stays a set of tests, and I’d sooner say that than dress it up.

The gap I’d bet on

The sharpest promise this brand owns is on the homepage, not in the ads. The site sells a sleep tracker built on not grading your night. The top of the ad account never says it.

Who buys a bedside sleep device? Rarely someone who sleeps well. It’s usually a person whose sleep has turned into something they worry about, and for them a nightly grade is the problem, not the feature. The site makes a second promise the ads skip: tracking from the nightstand, nothing worn, nothing to charge.

So the ads sell a category this buyer has often tried and quit, while the two reasons they might come back sit one click away.

Three concepts I’d test

The refusal. Open on the score, then take it away. The buying reason is relief from being measured, and it aims at the lapsed tracker owner. I’d kill it if the same audience converts better on a cut that keeps the score, because grading was then never the friction.

The swap. One object leaves the nightstand and another takes its place, one camera move, no voice over. The buying reason is getting the phone out of the bedroom, which the copy names but never shows. It fails when watch time drops before the swap lands, which would say the idea needs words after all.

The empty wrist. This one runs on the comparison the brand won’t make out loud, that it asks you to wear nothing and charge nothing. The buying reason is switching cost, aimed at people already inside the wearable habit. Dead on arrival if it pulls cheap attention and no adds to cart, the usual fate of comparison creative that argues instead of showing.

Each concept carries one claim, so a losing test tells you which claim died. Three variations of a single claim tell you nothing, and that’s how testing budgets disappear.

What I’d ship

Concepts and the test plan in two days. First cuts of all three inside a week, generated and not filmed, so a dead concept costs a day of work. Each version comes in 9:16, 4:5 and 1:1 with captions burned in, and the hook sits in a swappable layer, so round two is a data change, not a re-edit.

Where my limits are

I don’t run media buying and I don’t hold the account. I write and produce the creative, and I read results with whoever does the buying. I also can’t show past return numbers: the campaigns I’ve produced for were bought by the client’s own team and those numbers aren’t mine to publish. What I can show is the work and the speed of the loop.

On your account: same read, your name on it, and the parts I can’t see from outside marked unknown rather than guessed. Two days for the teardown and the test plan, a week for first cuts of all three concepts. [email protected] · the rest of the work sits on the studio page.