The single biggest reason accounts stall at scale isn't creative or audiences — it's structure. Too many ad sets, too little data in each, and an algorithm that never gets the signal it needs to find your buyers efficiently.
Meta's delivery system learns from conversion events. Every time you split your budget across another ad set, you split that learning too. Spread it thin enough and no single ad set ever exits the learning phase, so costs stay high and results stay volatile. The fix is almost always consolidation.
Start with fewer, bigger campaigns
For most accounts under roughly $50k/month, you don't need a sprawling map of campaigns. You need a small number of well-fed ones:
- One prospecting campaign using broad or lightly-stacked interest audiences, letting Meta's algorithm do the targeting work it's genuinely good at.
- One retargeting campaign catching warm traffic — site visitors, add-to-carts, engagers — with offers that close.
- A testing campaign, ring-fenced, where new concepts earn their place before they touch the budget that matters.
Every ad set you add divides your data. Every concept you consolidate multiplies it.
Feed the algorithm a clean signal
Consolidation only works if the event you're optimising for fires often and accurately. Aim for each ad set to clear roughly 50 conversions a week — that's the threshold where delivery stabilises. If you're well below it, optimise for an earlier event (add-to-cart, lead) until volume catches up, then move down the funnel.
This is also why broad targeting tends to win at scale: it gives the system the widest pool to find conversions in, which means more signal, faster learning, and lower costs once it settles.
Change one thing at a time
Structure is only stable if you stop poking it. Resist the urge to duplicate winning ad sets, nudge budgets daily, or restart learning every time a metric wobbles. Make one deliberate change, give it 3–4 days to settle, then read the result. Accounts that scale smoothly are almost always the calmest ones to manage.
What this looks like in practice
When we rebuilt WasteMates around this structure — consolidated prospecting, a tight retargeting campaign, and a separate testing lane — blended ROAS climbed to 3.84× while revenue grew 68%. None of that came from a clever hack. It came from giving the algorithm enough signal in one place to actually do its job.