At some point in the last 18 months, "creative is the new targeting" went from conference slide to reflex. Every media buyer can recite it. Far fewer have changed anything about how their ads actually get made. The monthly batch still looks like it did in 2023: one proven concept, eight near-identical cuts, a fresh hook bolted on the front, maybe a new colour grade. If that sounds like your account, the phrase has not yet reached the part of the operation it was aimed at.
The gap now costs real money. Meta rebuilt the machinery that decides who sees which ad, and the rebuild moved the audience-matching work out of your campaign settings and into the ads themselves. Your creative pipeline is your targeting stack. The question is what that changes on the first of next month, when the brief goes out.
What does "creative is the new targeting" actually mean?
The engineering story is short, and our full Andromeda explainer covers it properly, so here is the recap. Andromeda is the retrieval engine at the front of Meta's ad auction. Meta's engineering team announced it in December 2024: a system that sifts tens of millions of candidate ads in milliseconds, with roughly 10,000x the model capacity of what it replaced. Its job is to decide which ads are even plausible for a given person before the auction runs.
The part that matters for production: Andromeda makes that call by reading the ad. Who is on camera. What problem the first line names. The format, the setting, the claim, the tone. Those used to be creative decisions that came after targeting was set. They now are the targeting.
Meanwhile the old levers went soft. Interest stacks and lookalike ladders are treated as suggestions at best under Advantage+, and you cannot out-segment a model that builds its own segments from live behaviour. The one place you still decide who sees your ads is inside the creative itself.
Most teams hear this and conclude "make more ads". That reading is wrong, and it is the expensive kind of wrong because it feels like action. If your whole batch speaks to the same person about the same problem in the same shape, you have rebuilt a single narrow interest stack, however many files you upload. Meta's systems fold near-duplicates into one delivery entity. Volume without distinctness is one ad, uploaded 30 times.
Your media buyer used to decide who saw the ad. Now the brief does.
Where does the targeting live inside a UGC ad?
Three places: who delivers the ad, what problem it opens on, and what shape it takes. We pulled this apart in detail in our creative diversity piece, including the myth that 50 creators reading one script counts as diverse. The working version:
Persona is the strongest signal. A 26-year-old creator filming a get-ready-with-me speaks to a different corner of the auction than a 55-year-old talking about knee pain on a morning walk. Same product, different ad populations. Brief three genuinely different people with three genuinely different framings and you have opened three doors into the auction instead of one.
Angle is the second. The problem your opening line names tells the model what intent to match. "I stopped hitting snooze", "my sleep tracker finally went green" and "cheaper than my old routine" are three different audiences wearing the same product.
Format is the third and the most underrated. Talking head, voiceover B-roll, side-by-side demo, static with a review quote. Format changes who stops scrolling, and the system learns from who stops.
The test we apply to every batch before it ships: strip the logo and ask whether a stranger could sort the ads into separate "this one is for X" piles. If everything lands in one pile, you have variants, not diversity.
How to restructure your monthly creative brief
The operational change is smaller than the discourse suggests. You do not need a new team or new software. You need the brief to stop being a list of concepts and start being a grid.
- Rows are personas, not ideas. Pull 3 or 4 buyer types out of reviews, support tickets and post-purchase surveys. Real ones, described in the customers' own words, not marketing personas with stock-photo names.
- Give each persona 2 or 3 angles. Write the angle as the actual first line the viewer will hear, not as a theme. "Energy" is a theme. "I survived the 3pm crash without a third coffee" is an angle.
- Assign formats across the grid on purpose. Not every cell needs video. Mix talking head, voiceover B-roll, demo and static so neighbouring cells never share a shape.
- Cap variations of any winner at two. A winning creative earns one new hook and one new cut. It does not earn 15 crops of itself, because those get folded back into one entity anyway.
- Hold a quarter of the slots open. Fill them mid-month with iterations on whatever the live grid is telling you.
Here is what a 12-slot month looks like for a sleep supplement brand, with 9 committed cells and 3 held for mid-month iteration:
| Persona (from reviews) | Angle (opening line) | Format |
|---|---|---|
| New parent, 28 to 38 | "I survived the 3pm crash without a third coffee" | Creator talking head |
| New parent, 28 to 38 | "The wind-down routine that finally stuck" | Voiceover B-roll, evening routine |
| New parent, 28 to 38 | "Cheaper than the sleep stack I was buying" | Static, price breakdown + review quote |
| Shift worker, 25 to 45 | "Day sleep is a different sport. This is what worked" | Creator talking head, blackout-blind room |
| Shift worker, 25 to 45 | "What I take at 7am after a night shift" | POV demo, morning-after framing |
| Shift worker, 25 to 45 | "My partner sleeps 9 to 5. I don't" | Static, split-image with caption hook |
| Desk worker, 40+ | "I wake at 3am and my brain starts a meeting" | Older creator, talking head |
| Desk worker, 40+ | "My watch's sleep score, 30 days before and after" | Screen-record demo + voiceover |
| Desk worker, 40+ | "I tried magnesium, teas and podcasts first" | Voiceover B-roll, failed-alternatives montage |
| Open slots x3 | Iterate whichever cells the account is rewarding mid-month | Match the winning cell's format |
Each committed row is a distinct retrieval target. When results come back, read them by row, not by ad: which persona-and-angle cells found cheap buyers, which found nothing. Next month's grid doubles down on the live cells and swaps the dead ones. That reading habit compounds faster than any individual winner.
This is the pattern we see across Spark's own client accounts, where blended ROAS averages 8x. The accounts that hold that level share one habit: the brief is a persona-and-angle grid before a single script gets written. And when we take over an account that has been "doing creative volume" without the grid, the usual finding is 20 or 30 live ads that reduce to two actual signals. Same creator archetype, same pain point, different music. The account was testing edits, not audiences.
Key takeaway
Under Andromeda your monthly brief is a targeting document. Build it as a persona x angle grid, give every cell its own format, cap variations of any winner at two, and read results by cell rather than by ad. The metric that matters is distinct signals shipped per month, not files uploaded.
How much testing volume do you actually need?
More than most brands are comfortable with, less than the scare posts claim. Scaledon's analysis of post-Andromeda accounts, summarised in Segwise's 2026 playbook, found brands testing 20+ new ads a month running 65% higher ROAS than brands testing under 10, with the top third of advertisers holding roughly 395 live ads against 296 for the bottom third. Treat those as directional numbers from one dataset. They still match what we see in practice: accounts that keep feeding the system get cheaper, accounts that ration creative get repriced.
Two caveats before you triple your output. First, 20 ads only count as 20 if they are distinct. Twelve genuinely different persona, angle and format combinations will outperform 40 variants of one winner, because the variants collapse into a handful of signals. Second, the right number scales with spend. A $5k-a-month account does not need 40 ads; our testing quota guide sets the weekly numbers tier by tier from $5k to $100k+.
Fatigue is the other half of the volume question. Refresh windows that used to run six weeks now close in two to three, a pattern reported across post-Andromeda datasets and one we watched play out in home services accounts long before it had a name. Entire testing guides are now organised around this cycle; TheOptimizer's 2026 playbook is a decent example of the genre. The open slots in the grid exist precisely so the refresh never catches you flat.
The uncomfortable production truth sits underneath all of this. Sustaining 12 to 20 distinct creatives a month takes either a standing creator bench with an editing pipeline behind it, or a partner who runs one for you. That capacity question, not strategy, is where most teams actually get stuck; it is the exact problem our production system was built to remove.
What should you stop doing?
Every hour spent on the old levers is an hour not spent on the new one. The habits worth dropping first:
- Rebuilding interest stacks when performance dips. The model overrides them. A dip is almost always a creative signal problem now, not a settings problem.
- Running lookalike ladders. The 1%, 3%, 5% ad-set staircase splits your conversion signal and gives the algorithm less to learn from in each pot. Billo's write-up of the update lands where most of the credible coverage does: Andromeda rewards simpler structures, more volume and more distinct messaging.
- Duplicating ad sets to "reset" delivery. You are resetting learning, not luck.
- Cutting one winner into a dozen crops and calling it testing. That is a rendering exercise, not a test.
The hours come back somewhere useful. Review mining becomes a monthly ritual rather than a launch-week one, because the grid eats fresh angles. The creator bench becomes an asset you manage: 6 to 10 people who genuinely cover your personas, not one house style. And the weekly performance review moves up a level, from "which ad won" to "which cell won", which is the version of the question that tells you what to make next.
Creative is the new targeting: FAQs
What does "creative is the new targeting" mean on Meta?
It means Meta's delivery system, fronted by the Andromeda retrieval engine, decides who sees an ad by reading signals in the creative itself: who appears on camera, the problem named in the opening line, the format and the claim. Manual audience settings now mostly set guardrails. Distinct creatives reach distinct audiences, so the brief has replaced the audience tab as the place where targeting happens.
Should I stop setting audiences on Meta entirely?
No. Keep the guardrails: geography, language, hard age limits and any exclusions you need for compliance. Beyond those, broad targeting with Advantage+ defaults beats micro-segmentation in most accounts, because the creative is doing the segmentation. The hours you used to spend on interest stacks and lookalike ladders are better spent on the persona and angle grid in your brief.
How many new creatives should I test per month under Andromeda?
Roughly 10 to 20 distinct creatives per month at typical DTC spend, scaling with budget. One post-Andromeda dataset from Scaledon put brands testing 20+ new ads a month at 65% higher ROAS than brands testing under 10. Distinctness matters more than the count: 12 different persona, angle and format combinations beat 40 variants of one winner.
Is 20 variations of a winning ad enough creative diversity?
No. Meta's systems group near-duplicates and treat them as one creative signal, so 20 crops of the same winner reach roughly the same audience as the original ad. Cap variations of any winner at two (a new hook and a new cut) and spend the remaining production budget on new persona and angle combinations that open new pockets of the auction.
"Creative is the new targeting" is a production sentence wearing a media-buying costume. The teams pulling ahead this year are not sharper in Ads Manager. They run better grids, feed them every month, and treat the brief as the most important targeting document in the account. If you want the delivery model and costs behind running that at volume, our pricing page lays it out, and the wider resources library covers the rest of the system.