The Algorithm Needs Volume, Not Instructions
Meta's delivery system learns from conversions. Each ad set goes through a learning phase, and Meta's own guidance is that an ad set needs roughly 50 optimisation events within a week to stabilise. Split a ₹2L monthly budget across twelve ad sets and most of them will never get there. They stay in learning, costs stay volatile, and the account looks like it "doesn't work."
The fix is not a smarter interest. It is fewer, larger buckets of budget that give the system enough signal to optimise.
"On Meta, every extra ad set is a tax on learning. Consolidate until each one can realistically hit its weekly conversion volume."
A Structure That Scales
For most e-commerce and lead-generation accounts, three campaigns do the heavy lifting:
1. Prospecting (the main engine)
One campaign, campaign-level budget (what Meta now calls Advantage campaign budget, previously CBO), and broad targeting — country, age floor if needed, and little else. Your creative does the targeting: an ad about postpartum skincare finds new mothers far more reliably than an interest called "parenting."
2. Creative testing
A smaller, separate campaign (typically 10–20% of spend) where new concepts are tested against each other. Winners graduate into the prospecting campaign. This keeps experiments from destabilising the ads that are already paying the bills.
3. Retargeting (only if it earns its place)
Broad prospecting already reaches warm audiences, so a dedicated retargeting campaign often just re-buys sales you would have got anyway. Keep it small, use it for specific offers or reminders, and judge it on incremental results rather than platform-reported ROAS.
Structure Checklist
- Can every ad set realistically reach ~50 conversions a week? If not, consolidate.
- Is targeting broad enough for the system to find buyers, with creative doing the segmentation?
- Is new creative tested in its own campaign instead of inside your best performer?
- Are you optimising for the event that matters (purchase or qualified lead), not clicks or landing page views?
- Is your Pixel backed by the Conversions API so the system sees conversions that browsers miss?
Stop Editing Every Day
Significant edits — big budget jumps, new targeting, swapping the optimisation event — can push an ad set back into learning. Daily tinkering keeps an account permanently unstable. We review Meta accounts on a fixed weekly rhythm, scale budgets in measured steps, and make structural changes deliberately, with a note in the action log explaining why.
What to Optimise For
The algorithm will find exactly what you ask for. Optimise for link clicks and you get clickers. For e-commerce, optimise for purchases (or add-to-cart only while purchase volume is too thin). For lead generation, optimise for the lead event — and wherever possible feed back qualified leads from your CRM so Meta learns which leads actually turn into revenue.
Measuring It Honestly
Meta's reported ROAS is useful for comparing ads against each other, but it overlaps with every other channel's attribution. For business decisions, look at blended numbers — total revenue against total marketing spend — and watch whether new-customer volume rises when Meta spend rises. Our article on ROAS, MER and CAC goes deeper.
Simple structures are harder to build than complicated ones, because they force you to make decisions. But they give the algorithm what it needs, they make results easier to read, and they free your time for the lever that matters most on Meta: creative.