Favour consolidation over fragmentation. Meta’s delivery system optimises best when campaigns have room to learn, so a few well fed campaigns usually outperform many thin ones. Keep testing separate from scaling, use a naming convention you can report on, and let the system do the audience work it is built to do.
The instinct that wrecks most DTC accounts is control, so name it early and resist it. Marketers coming from older platforms want to slice everything into tight little ad sets, one per audience, one per interest, one per placement. On today’s Meta, that instinct works against you. The delivery system is a machine that needs volume to learn, and fragmentation starves it. Structure the account so the machine can do its job, and get out of its way.
Every ad set on Meta has to pass through a learning phase, during which the system gathers enough conversion events to find the people most likely to buy. That phase needs a meaningful volume of events to stabilise. When you fragment a budget across many ad sets, each one gets a trickle of spend, none of them accumulates enough events to exit learning cleanly, and the system never gets a confident signal to optimise against. You end up with a dozen under fed ad sets all guessing.
Consolidation fixes the input problem. Fewer campaigns and fewer ad sets, each carrying more budget and more conversion volume, give the delivery system the density of signal it needs to learn and then optimise. You also stop competing against yourself, which is what happens when several of your own ad sets chase overlapping audiences in the same auction and bid up your own costs.
Consolidation is not the same as carelessness. It means deliberately choosing the smallest number of campaigns that still gives you clean, separate reads on the decisions you care about, then feeding each one properly. The question to ask of any proposed split is simple: does this division give me a decision I will actually act on, or am I just slicing for the comfort of seeing more rows in the reporting table?
Meta’s targeting has moved steadily toward broad. The system is now generally better at finding your buyers across a large or open audience than a marketer is at hand picking narrow interests, because it optimises against real conversion behaviour rather than a guess about who your customer is. Practically, that means broad audiences give the delivery system the widest pool to learn from, and heavy manual segmentation often just fragments that pool and slows learning without adding real precision.
The creative, not the audience layer, is increasingly where the actual targeting happens. A hook aimed at new mothers finds new mothers. A message about a specific pain point self selects the people who feel that pain. On a broad, well fed setup, your creative does the pointing, and the system does the finding. That shift is the single biggest reason old fragmentation habits underperform today.
The one separation worth protecting is between testing and scaling, because they have different jobs and different definitions of success.
Testing is about learning. Its purpose is to find the creative angles, hooks and offers that resonate, cheaply and quickly, before you commit real budget. A testing structure is built to give new concepts a fair, isolated read without disrupting the campaigns that are already carrying your revenue. This is where a steady stream of native testing content earns its keep.
Scaling is about exploitation. Once a concept proves itself in testing, it graduates into a scaling structure whose job is to spend efficiently against what already works, with stability and enough budget to keep learning healthy. Mixing the two is where accounts get chaotic: shoving an unproven idea straight into your main scaling campaign risks resetting its learning and disturbing performance that was paying the bills, while babysitting proven winners in a testing setup caps their reach.
Keeping the two separate gives you a clean pipeline. Ideas are proven in a low stakes environment, winners graduate to a high stakes one, and each structure is judged on the thing it is actually there to do. The exact mechanics of promotion, meaning what threshold a concept must clear before it graduates, should be defined against your own volume and economics, and are worth setting deliberately rather than by feel.
None of the structure above is useful if you cannot read it. A consistent naming convention, applied from day one across campaigns, ad sets and ads, is what turns a pile of spend into a report you can actually analyse. When names encode the things you want to slice by, such as concept, angle, format and stage, you can pull performance by any of those dimensions in seconds. When names are ad hoc, every analysis becomes an archaeology project.
Measurement deserves the same discipline. Decide up front what you are optimising for and how you will judge it, because Meta’s in platform reporting and your own blended, post purchase view will rarely agree. In platform numbers are shaped by the attribution window and the modelling the system uses. Your true picture usually comes from combining that with a business level view of blended return and new customer economics. Structure the account so your key decisions map to numbers you trust, not just the ones the interface shows you by default.
A sane DTC Meta structure is really one idea applied repeatedly: give the delivery system dense signal and clear jobs. Consolidate so campaigns can learn. Go broad so the system can find buyers and let creative do the pointing. Separate testing from scaling so each has one job. Name and measure so you can read what happened and decide what to do next. Get those four right and the account becomes legible, which is the precondition for scaling it with any confidence.
Consolidate, in most DTC cases. Meta’s delivery system needs a meaningful volume of conversion events to exit the learning phase and optimise. Fragmenting budget across many thin ad sets starves each one of signal, so a few well fed campaigns usually outperform many small ones.
Broad audiences generally give the delivery system the widest pool to learn from, and it is now often better at finding buyers than manual interest picking. On a broad setup, your creative does the targeting: the hook and message self select the right people.
Because they have different jobs. Testing is for cheaply proving new creative and offers in isolation. Scaling is for spending efficiently against proven winners with stability. Mixing them risks resetting learning on your revenue campaigns and disturbing performance that already works.
There is no fixed number. The guiding rule is the smallest set of campaigns that still gives clean, separate reads on decisions you will act on, with each campaign fed enough budget and conversion volume to learn. Add a split only when it yields a decision.
Because it decides whether you can analyse your account at all. Consistent names that encode concept, angle, format and stage let you pull performance by any dimension instantly. Ad hoc names turn every analysis into a manual reconstruction.
We will audit the structure, consolidate what is starved, and separate testing from scaling.