Question 1

What is the smallest Meta budget that can actually learn?

Accepted Answer

Work it backwards from documented platform behaviour. Meta states an ad set typically needs about 50 optimisation events inside a week to exit the learning phase. At an illustrative ₹290 cost per lead, 50 conversions cost 50 x 290 = ₹14,500 per ad set per week - call it about ₹62,000 a month for one ad set that genuinely learns. Now invert it: a ₹29,000 weekly budget funds two learning ad sets, not eight starving ones. Split the same money eight ways and each ad set collects roughly 12 conversions a week, sits in learning forever, and posts volatile costs that read worse than they should. The arithmetic is illustrative - your own cost per lead sets the exact bar - but the law it points to is fixed: size the structure to the signal, not to the org chart.

Question 2

Why are our Meta leads such poor quality?

Accepted Answer

Usually because the form asks for nothing. A name-and-number instant form is frictionless by design, which means it collects everyone - including people who tapped by accident. We rebuild the form with qualifying questions so budget, timeline or intent is captured before the lead ever reaches your sales desk. The second cause is the optimisation signal. If Meta is optimising to form fills rather than qualified leads, it will keep finding more of the same low-intent traffic. Wiring the CRM back into the platform fixes the signal at the source.

Question 3

Do Meta ads still work after the Apple privacy changes?

Accepted Answer

Yes, but measurement had to be rebuilt. The browser Pixel alone lost visibility when Apple restricted tracking, so we run the Conversions API server-side alongside it, deduplicated on a shared event ID so nothing counts twice - the Dedup Spine. That pairing recovers a large share of the events the browser no longer reports. On top of that, CRM events flow back into the platform so Meta optimises to leads that actually qualified, not just form opens. That is the change that keeps Advantage+ and your Lookalikes pointed at buyers rather than volume.

Question 4

Should we run Advantage+ Shopping from day one?

Accepted Answer

Only if the catalogue and the conversion volume are ready for it. Advantage+ Shopping folds prospecting and retargeting into one heavily automated campaign, and it needs enough conversion signal to learn - handing it spend too early just burns budget while it guesses. Until then we run structured campaigns with named creative themes and clean exclusions, then graduate winners into Advantage+ once the volume supports the handover. Automation multiplies good signal and taxes bad signal.

Question 5

CBO or ABO - which should our account use?

Accepted Answer

Campaign budget optimisation (Meta now files it under Advantage campaign budget) lets the platform move spend across ad sets toward whatever converts cheapest right now, which is what you want on proven structure at scale. Ad set budgets guarantee each cell its spend, which is what you want when a test needs a fair read. Our default is ABO for tests, CBO for scaling, and never both jobs in one campaign. Most messy accounts we audit are trying to test and scale in the same place, so neither happens well.

Question 6

Why does the learning phase matter so much?

Accepted Answer

Because an ad set in learning shows you noise, not numbers. Below the documented bar - about 50 optimisation events in a week - delivery stays volatile and costs read higher than they should, which is why the budget math in the first question decides structure before anything else does. It is also Law 04: respect the learning phase. Significant edits - budget jumps, audience swaps, new creative - ship in scoped batches, because each one can reset the clock, and an account that is permanently learning never shows you its real performance.

Question 7

Broad targeting or interest stacks - which works now?

Accepted Answer

Both can, and the account decides. Broad audiences give the delivery system room to find buyers when creative is strong and the conversion signal is clean; interest stacks and Lookalikes still earn their place where budgets are small or the product is narrow. What stays constant is the discipline: one variable tested at a time, exclusion audiences on so customers and open deals never see prospecting spend, and Lookalike seeds built from qualified pipeline rather than raw form fills.

Question 8

How often should creative actually change?

Accepted Answer

On a planned cadence, not after performance sags. Our standard is themed refreshes every 14 days, with hook rate and frequency watched weekly - rising frequency with falling CTR is the Creative Fatigue Curve announcing itself, and fatigue is cheapest to fix while it is still a slope, not a cliff. Each theme ships with a written hypothesis and a kill criterion. Winners are logged so the next round builds on evidence; losers retire fast instead of dragging the account average down.