Structure is the quietest decision in media buying and one of the most consequential. How you split your campaigns, ad sets and ads decides what the algorithm can learn, how cleanly you can read a result, and whether scaling a winner is a two-click move or a rebuild. Most accounts that plateau are not short of good creative — they are drowning it in a messy structure that never lets a winner surface.
Every major paid platform shares the same three-level shape: a campaign that sets the objective, one or more ad sets that control targeting and budget, and the ads themselves that carry the creative. Get the roles of those three levels right and everything downstream — testing, budgeting, optimisation — gets easier. This guide walks the hierarchy, the one-concept-per-ad-set rule, the ABO versus CBO decision, where Advantage+ style automation fits in 2025-26, and the naming discipline that keeps a growing account readable. If you are new to the discipline, start with what media buyers actually do and come back.
Think of the three levels as answering three different questions. The campaign answers what outcome am I buying — a purchase, a lead, an install, an app event. The ad set answers who am I showing it to and how much am I willing to spend — the audience, placements, budget and bid strategy live here. The ad answers what do they actually see — the hook, the visual, the copy, the landing page. Keep those jobs separate in your head, because mixing them is where structure goes wrong. When people say a campaign is "messy," they almost always mean targeting decisions and creative decisions have been tangled together so no single change can be measured on its own.
The practical rule that falls out of this: change one variable per level at a time. If you are testing audiences, hold the creative constant across ad sets. If you are testing creative, hold the audience constant and vary the ads. When two levels move at once, a lift in results has two possible parents and you will never know which one to scale.
The single most useful structural habit is one concept per ad set. A concept is a distinct idea about why a user should care — an angle, a promise, a proof point. When each ad set carries one concept expressed through a few close variations, the delivery system can pool learning around that idea and give you a clean verdict on whether the idea works. Cram five unrelated angles into one ad set and the algorithm spends your budget picking a favourite before you have enough data to trust the pick, and the losers quietly starve the test.
This is why the structure and the creative testing framework are really one topic. Structure is the container; testing is what you pour into it. A disciplined tester runs each new angle in its own ad set, keeps two to four executions of that angle inside it, and reads the ad set as a unit. Winners get promoted; the concept, not just a single clip, is what proved itself.
The biggest structural fork is where you set the budget. With ABO (ad-set budget optimisation) you set a budget on each ad set, so every audience or concept gets a guaranteed, equal shot regardless of early results. With CBO (campaign budget optimisation, now surfaced under Advantage names on Meta) you set one budget at the campaign level and let the algorithm push spend toward whichever ad set is performing in the moment. Neither is "correct" — they serve different jobs, and experienced buyers use both.
The clean division of labour: ABO for testing, CBO for scaling. When you are testing, you want each concept to reach its own event threshold without a fast starter stealing the budget before a slower concept has proven itself, so ABO's guaranteed split is the honest way to run it. Once testing has told you which concepts and audiences win, you bundle those proven performers into a CBO campaign and let the algorithm allocate in real time. Meta's own guidance leans toward CBO for most scaling campaigns and suggests roughly three to five ad sets inside one, which keeps the pool broad enough to optimise without fragmenting delivery.
| Dimension | ABO (ad-set budget) | CBO (campaign budget) |
|---|---|---|
| Budget set at | Each ad set | Campaign, split by algorithm |
| Control | High — you guarantee the split | Lower — algorithm decides |
| Best job | Testing & fair comparison | Scaling proven winners |
| Risk | You fund losers longer | Winner-take-all starves valid tests |
| Reading results | Clean, per-concept | Noisier, campaign-level |
The third option in 2025-26 is fully automated campaigns — Meta's Advantage+ family being the best known — where the platform folds audience, budget and creative selection into one product. You load many ads into a single campaign, hand over a broad audience and one budget, and the system decides the rest. On offers with strong signal and a broad addressable market this can outperform hand-built structures, because modern retrieval engines evaluate far more combinations than a human can.
It is not a default, though, and the market has been ambivalent. Industry reporting suggests Advantage+ style automated spend pulled back through late 2025 into 2026 as advertisers took manual control back to protect testing discipline — one widely cited estimate put its share of retail spend around 20% in early 2026, down from a peak near 38% a year earlier (these figures come from third-party analyses rather than platform disclosures, so treat the exact numbers as directional). The operator takeaway is steady regardless of the percentages: automation is a powerful scaling tool once you know what wins, but it hides the very signal you need while you are still learning. Use structured ABO to find winners, then decide whether automation or a hand-built CBO scales them best. The broader shift toward feeding algorithms rather than micromanaging settings is covered in AI for media buyers.
A naming convention feels like bureaucracy until the day you have forty ad sets and cannot tell what any of them tests. A good name encodes the variables you will later want to filter and report on, in a fixed order, so the account reads like a spreadsheet. A workable pattern is objective, audience, concept, format, date — for example PUR_broad_price-angle_ugc_0725. The exact scheme matters far less than picking one and never breaking it, because consistency is what lets you slice performance by concept or format later without opening each ad.
The same discipline applies to your creative library and your tracking. When ad names, creative filenames and your tracking parameters all carry the same concept label, you can follow one idea from the ad account through to conversions without guesswork. Sloppy naming is a tax you pay every single time you try to read the account.
Modern delivery is a matching engine that needs enough conversions per unit to learn. Every ad set has to accumulate a minimum volume of the event you optimise for — commonly cited as roughly 50 conversions in a week — before its delivery stabilises out of the learning phase. Structure decides whether that volume is reachable. Split one budget across ten thin ad sets and none of them ever gathers enough signal; consolidate into three well-fed ad sets and each can actually learn. Fragmentation is the most common self-inflicted wound in a media buying account.
This is why "more ad sets" is rarely the answer to poor performance. Each new ad set divides your signal; each duplicated audience competes with itself in the same auction. The structural instinct that pays off is consolidation — fewer, better-fed units carrying clearly separated concepts. Get the container right and budgeting and optimisation become tractable, which is exactly what budget allocation and the campaign optimisation workflow build on top of. Scaling a clean structure is a promotion; scaling a messy one is a rebuild — see scaling campaigns.
Fewer than instinct suggests. For a scaling CBO campaign, three to five well-funded ad sets is a common sweet spot — enough variety for the algorithm to optimise across, few enough that each still gathers the conversion volume it needs to exit the learning phase. When you are testing on ABO you may run more, but each still needs its own budget to reach a fair verdict.
Test on ABO. Guaranteeing each concept its own budget is the only way to compare ideas fairly, because CBO will divert spend to an early leader before a slower-starting concept has had a chance to prove itself. Move winners into CBO once you are scaling and want the algorithm allocating in real time.
It depends on the stage. Automated campaigns can outperform hand-built ones when scaling a proven, broad-appeal offer, because they evaluate more combinations than a person can. While you are still learning what wins, though, automation hides the per-concept signal you need, so most operators find winners in a structured setup first and only then hand them to automation.
Usually because edits are re-triggering the learning phase or because budgets are split so thin that no ad set ever gathers enough conversions to stabilise. Consolidate into fewer, better-funded ad sets, make significant edits sparingly, and give each unit the volume it needs before you judge it.
What media buyers actually do day to day: testing creatives, reading data, managing spend and scaling the campaigns that turn a profit.
Advanced · 14 min readThe algorithm handles targeting now; creative is what you control.
Advanced · 12 min readThe angle is the highest-leverage decision in a campaign — the specific reason THIS audience wants THIS offer.
