Set up your first Meta campaign

~2,000 words · campaign settings top to bottom · post-Andromeda playbook for a brand-new ad account

The premise here is the current best practice on Meta: you're not testing audiences anymore, you're testing persona × awareness-stage combinations. Andromeda has gotten good enough that hand-picked interest stacks usually cost you performance. Your creative is your targeting. So the structure isolates one persona per ad set, gives each a small clean budget on broad targeting, and stocks the ad set with four creatives spanning the full T-E-E-P funnel, so Andromeda can route the right awareness-stage creative to each buyer's psychological state.

Older Meta playbooks, anything from before late 2025, will tell you to build interest stacks, layer lookalikes, and exclude targeting categories. Ignore that, and ignore most of Meta's in-platform recommendations too. They tend to push more spend, not more profit.

Naming Conventions: Set These Before You Build Anything

Boring but load-bearing. Consistent naming is what lets you read your account at a glance two months from now, and what makes reporting and UTM-based attribution work cleanly. Use these exact formats.

Level	Format	Example
Campaign	Testing	Testing
Ad Set	(Audience descriptor) Name of Persona	(Broad US 25-54) Dewy Skin Seeker (Broad US 25-54) Sensitive Skin Sufferer
Ad	Persona Name + Stage + Ad Type + Number	Dewy Skin Seeker T Video 1 (Trigger) Dewy Skin Seeker E1 Video 1 (Exploration) Dewy Skin Seeker E2 Image 1 (Evaluation) Dewy Skin Seeker P Image 1 (Purchase)

The Testing Campaign: Settings, Top To Bottom

Level	Setting	What to choose
Campaign	Buying type	Auction
Campaign	Objective	Sales
Campaign	Advantage+ vs Manual	Manual while testing. Advantage+ Shopping (ASC) automates away the per-persona signal you're trying to read, so save it for scaling.
Campaign	Campaign budget optimization (CBO)	Off. Budget lives at the ad set, not the campaign.
Campaign	Budget scheduling	Off
Campaign	Catalog	Off
Ad Set	Budget	$10/day per ad set (ad set budget optimization, ABO)
Ad Set	Budget sharing	Off
Ad Set	Ad scheduling	Off
Ad Set	Conversion location	Website
Ad Set	Conversion event	Purchase (not View Content or Add to Cart)
Ad Set	Optimization	Highest Volume
Ad Set	Attribution	7-day click / 1-day view (default). Drop the 1-day view if your view-through > 20%.
Ad Set	Audience	Broad. Country, broad age range (e.g. 25-54), gender if clearly relevant. Do NOT let Meta control audience targeting. Do NOT stack interests or lookalikes. Exclude purchasers.
Ad Set	Placements	Automatic placements (Advantage+ placements); this one Meta default is fine
Ad Set	Languages	All
Ad Set	Dynamic Creative	Off (we're testing angles, not asset combos)
Ad	Format	Mixed: video + static + carousel where the angle calls for it

Why $10/Day Per Ad Set

$10/day is a common starting point for early angle testing: enough delivery over a few days to read creative and persona signal, and small enough to run several ad sets in parallel. It usually won't fully exit Meta's learning phase, which needs many more conversions, but you're reading directional signal here, not chasing statistical significance. At $10/day Meta also spreads delivery across placements over a few days, a more honest read than a $50/day spike.

The Testing Campaign, Visualized

📦 Campaign: Testing
   ├── Sales objective
   ├── Auction buying
   ├── Manual setup, CBO off
   │
   ├── 🎯 Ad Set: (Broad US 25-54) Dewy Skin Seeker  ·  $10/day
   │     Broad targeting, exclude purchasers
   │     ├── T   "It's 11pm and your skin looks like 11pm"          (problem-aware scene)
   │     ├── E1  "5 products = the glow you keep saving on IG"      (solution-aware)
   │     ├── E2  Before/after 30-day testimonial UGC                (product-aware proof)
   │     └── P   "Bundle + 30-day guarantee, free returns"          (most-aware offer)
   │
   ├── 🎯 Ad Set: (Broad US 25-54) Flake Survivor  ·  $10/day
   │     Broad targeting, exclude purchasers
   │     ├── T   "My foundation pills by 2pm every day"             (problem-aware scene)
   │     ├── E1  "It's not your makeup, it's the dryness underneath" (solution-aware)
   │     ├── E2  Founder explains the 2,700 ingredients removed      (product-aware)
   │     └── P   "Risk-free trial, send it back if it doesn't work"  (most-aware offer)
   │
   └── 🎯 Ad Set: (Broad US 25-54) Sensitive Skin Sufferer  ·  $10/day
         Broad targeting, exclude purchasers
         ├── T   "Every new product = burning. Sound familiar?"     (problem-aware scene)
         ├── E1  "Clean beauty ≠ clean ingredients. Here's why."    (solution-aware)
         ├── E2  Dermatologist UGC review, rating overlays          (product-aware proof)
         └── P   "Skip the patch test. We did 6 months of them."    (most-aware offer)

One persona = one ad set. Inside each ad set, four creatives spanning Trigger → Exploration → Evaluation → Purchase (Sarah Levinger's T-E-E-P framework, covered in detail in Studying what wins on Meta now). The ad set tells you "does this persona convert?"; the spend pattern across the four creatives tells you "what awareness stage is this persona actually in?" Don't mix personas inside a single ad set; that's the line you can't cross. But do mix awareness stages within a persona, which is what Andromeda needs to route the right ad to each buyer state.

Creative Size Specs

Don't waste a day's testing on creatives that get auto-cropped wrong or downranked for resolution. The current Meta specs:

Format	Aspect ratio	Size	Where it shows
Single Image, Feed (Square)	1:1	1080 × 1080 px	Facebook Feed, Instagram Feed
Single Image, Feed (Vertical)	4:5	1080 × 1350 px	Mobile Feed (preferred, takes more screen real estate)
Video, Square	1:1	1080 × 1080 px minimum	Feed
Video, Vertical	4:5 or 9:16	1080 × 1350 / 1080 × 1920	Reels, Stories, mobile Feed

Meta technically accepts video as small as 600 px wide. Don't ship at 600 px. Ship at 1080 minimum.

What To Read Once The Data Comes In

After 3-5 days at $10/day per ad set, you'll have meaningful data. The richer structure means you now read both audience signal and awareness-stage signal in the same window. Questions in order:

1. Which ad sets are hitting your KPIs? At this budget you're reading directional signal, so weigh the leading indicators together: cost per purchase and ROAS where you have purchases, plus link CTR and cost per click. Keep every ad set that's hitting goal; those are your winning personas. If they're all failing, kill the whole test, take the most positive signals you saw, and build another iteration from those.
2. Inside a winning ad set, where did the spend land across T-E-E-P? Andromeda routes by efficiency, so the stage that absorbed the most spend usually signals where your buyers are in their journey. This is often the most valuable read in the whole campaign.
3. Inside a winning ad set, which specific creative did the most converting? That's your winning awareness-stage × format combination, the asset to multiply when you scale.

Some patterns to recognize:

What the spend pattern looks like	What it means	What to do next
~80% of spend went to the T (Trigger) creative, healthy CPA	Your buyers are pre-awareness. They don't yet recognize the problem. The brand's job is awareness, not conversion.	Multiply more Trigger creatives. Hold off on offer-led ads.
~80% of spend went to E1 or E2 (Exploration / Evaluation), healthy CPA	Buyers are aware of the problem and are actively comparing options. You're competing against alternatives.	Multiply proof-heavy creative: testimonials, before/afters, founder explainers.
~80% of spend went to P (Purchase), healthy CPA	Buyers are ready to buy. Friction is the only thing in the way.	Sharpen the offer. Test guarantees, bundles, free shipping thresholds.
Spend distributed evenly across all four, healthy CPA	Strongest possible signal; your concept works across the full funnel.	Scale aggressively. Build out the Scaling campaign with more of this persona across more formats.
Spend concentrated on one stage, poor CPA	Persona may be right but the creative for the dominant stage is weak.	Don't kill the persona yet. Replace the dominant-stage creative and re-test.
Spend evenly spread, poor CPA across all four	The persona is probably wrong. Andromeda isn't finding efficient delivery to anyone.	Kill the ad set. Go back to research for a different persona.

When To Kill

The kill rule: once an ad set has spent about 1.5-2× your target cost per purchase with zero sales and link CTR is running at roughly half your target or less, kill it. So if your target cost per purchase is $20, that's about $30-40 of spend with nothing to show; if it's $60, you'd wait until $90-120. The persona was wrong, or the T-E-E-P creative mix was weak across the board, or both. Move budget to the ad set that's working and pull a new persona from your icebox. This is the same goal-relative threshold used in Tips for pausing, fixing, or scaling ads; fixed dollar amounts don't travel across accounts.

If link CTR is fine but conversions aren't happening, the problem is usually downstream: landing page, offer, or checkout. Fixing the creative won't move it.

If You Remember Nothing Else

Three ideas worth carrying out of this series:

• Dara Denney: Start from a diagnosis. A clear, one-sentence read on what's actually happening with your buyer is what turns scattered week-to-week tests into a real strategy.
• Will Sartorius: The AI creative pipeline is simpler than it looks. It's a set of text files you build once and reuse, so it's worth getting comfortable with.
• Meta, in 2026: Your creative is your targeting. Speak specifically to one person, give Andromeda broad surface area, and let it find more people like them.

That's it. Go make ads.

Sources and credits. The current best-practice campaign structure synthesized here draws on the public work of Sarah Levinger (T-E-E-P framework), Dara Denney, Partner at Point Guard Media (research-to-strategy flow, the diagnosis concept), and Will Sartorius, CEO of SelfMade / Skipper (AI creative pipeline). All three publish openly across YouTube, LinkedIn, and their own newsletters; go follow them, that's where the real-time updates live. The campaign launch structure, the post-Andromeda "one persona per ad set × four T-E-E-P creatives" framing, and the starter-vs-scale guidance are ours. Andromeda details from Meta Engineering blog (Dec 2024) and subsequent rollout coverage. If the frameworks land, that's because of Sarah, Dara, and Will. If anything is wrong, that's on us.