Meta Ads Account Structure: Simpler Beats Clever
Meta's delivery system has quietly become the strategist. Most accounts underperform not because the creative is bad but because the structure gives the algorithm too little data in too many places to learn anything.
Understand what fragmentation costs you
Every ad set needs roughly 50 conversions per week to exit the learning phase and optimise properly. An account with twelve ad sets and 120 weekly conversions gives each one ten — none of them ever learn, and all of them deliver at learning-phase efficiency.
The same 120 conversions in two ad sets gives each sixty. Identical budget, identical creative, materially better delivery. Consolidation is usually the highest-return change available in a small account.
Use broad targeting and let the creative do the targeting
Detailed interest stacks made sense when the algorithm needed help. It no longer does, and narrow audiences now mainly raise costs by shrinking the pool the system can optimise within.
Broad targeting with the country and an age range is the sensible default for most accounts. Your creative becomes the targeting mechanism: the hook selects the audience far more precisely than an interest checkbox ever did.
A structure that works for most accounts
Unless you have a specific reason to deviate, start here and only add complexity when data forces you to:
- One prospecting campaign with campaign budget optimisation, one or two broad ad sets, and four to six creatives per ad set.
- One retargeting campaign covering site visitors and engagers over 30 days, with different creative — not the same ads at higher frequency.
- One testing campaign at a small fixed budget, where new creative angles run before graduating into prospecting.
- Nothing else until a specific question requires its own campaign.
Set budgets where you want control
Campaign budget optimisation lets Meta move money to whichever ad set is performing, which is usually what you want. Ad set budgets are for when you need to guarantee spend somewhere — a new market, a product line the business has committed to regardless of efficiency.
Be careful with budget edits. Large changes reset the learning phase, so move in increments of roughly 20% and wait several days between changes rather than adjusting daily.
Name things so reporting is possible later
Naming feels bureaucratic until you are three months in and cannot answer which creative angle works. Use a consistent pattern that encodes what you will want to filter by: objective, audience, angle, format, date.
Something like PROSP_BROAD_FOUNDERSTORY_VIDEO_2026-04 is ugly and genuinely useful. Without it, every analysis becomes archaeology.
Frequently asked
How many creatives should I run per ad set?
Four to six active at a time. Fewer and Meta cannot find a winner; more and the budget splits so thin that nothing accumulates enough data. Replace the weakest one or two each week rather than rebuilding the whole set.
Should I still use lookalike audiences?
They still work in larger accounts with a strong source list, but the gap over broad targeting has narrowed considerably. Test them against broad rather than assuming they win — in most accounts under ₹5 lakh monthly spend, broad now matches or beats them.
How long should I wait before judging a campaign?
Until it exits the learning phase and has accumulated a meaningful sample — typically seven to fourteen days, or about fifty conversions per ad set. Judging on day three reliably kills campaigns that would have worked.