The exhaustive Meta Ads FAQ.

Enough for your ad sets to exit the learning phase — Meta wants roughly 50 optimisation events per ad set per week. If your target CPA is $20, that's ~$140/day across the account to learn properly. Underfunding many ad sets is worse than funding one well.

No — both run from one Meta campaign, and Advantage+ placements will spread across both automatically. Let delivery find where your audience responds rather than guessing; you can always exclude a placement that spends without converting.

After launch or a significant edit, Meta spends a period (roughly 50 conversions per ad set) calibrating who to show your ad to. Performance is volatile during it, and every big edit resets it. Constant tinkering keeps you permanently in learning — the most common self-inflicted wound.

Boosting is a simplified shortcut with limited objectives and targeting — fine for a little local visibility, wrong for real results. Proper campaigns in Ads Manager give you conversion objectives, real audiences and actual measurement.

CBO (campaign budget) lets Meta shift budget between ad sets automatically; ABO (ad set budget) gives you fixed control per ad set. CBO works well with a few proven ad sets; ABO is better for testing, where CBO would starve unproven ad sets of the data they need to learn.

Fewer, better-funded ad sets beat many starved ones — consolidation helps learning. A common shape: 1–3 ad sets per campaign, 3–6 creatives per ad set, so Meta has variety to test without splitting data too thin.

Meta's heavily-automated shopping campaign — minimal targeting input, broad delivery, Meta decides almost everything. Often strong for ecommerce with good tracking and creative volume, but it's opaque: like Google's PMax, it rewards oversight it doesn't natively provide.

Scale winners gradually — big jumps reset learning. A common rule is increasing budget ~20% every few days rather than doubling overnight. Vertical scaling (more budget) and horizontal scaling (new audiences/creatives) are both options; the second is often more durable.

A Custom Audience is people you already know — your customer list, site visitors, engagers. A Lookalike is Meta finding new people similar to that source. Customs are for retargeting; Lookalikes are for prospecting. The Lookalike is only as good as the source you build it from.

Bigger than instinct suggests — Meta's delivery works best with room to optimise, and hyper-narrow interest stacking mostly just raises costs now. Broad targeting with strong creative frequently beats micro-targeting; the creative is the targeting on today's Meta.

Less each year. Meta has removed thousands of interest options and its delivery increasingly ignores narrow inputs (Advantage+ audience treats your targeting as a suggestion). Effort is better spent on creative variety and clean conversion signals than interest-stacking.

Showing ads to people who already visited, engaged, or bought — the warmest traffic you have. It's usually the cheapest conversions in the account. But it needs accurate tracking to populate the audiences, and it can't scale beyond the traffic you feed it.

Native-feeling beats polished — content that looks like it belongs in the feed. Strong first 1–3 seconds (the hook), vertical video for Reels/Stories, and a clear single message. What works is account-specific: test angles, then double down on winners.

An audience has seen your ad too many times: frequency climbs, CTR falls, CPA rises. The fix is refreshing creative before the decline shows up in results — which means watching the early signals (frequency + CTR trend), not waiting for the monthly report.

More often than most businesses plan for — winning ads burn out in weeks, faster with small audiences and big budgets. Sustainable accounts have a steady pipeline of variants generated from what's already converting, not a scramble after performance drops.

The first 1–3 seconds of a video ad (or the first line of copy) — the part that decides whether anyone sees the rest. Different hooks on the same ad can produce wildly different results, which is why testing hooks specifically, not just whole ads, pays off.

Start automatic (Advantage+) — Meta shifts spend to what's working and cheap placements subsidise expensive ones. But audit where spend actually goes: if a placement has meaningful spend and zero conversions, exclude it deliberately.

Meta showing your ads inside third-party apps and sites. It's cheap reach, and sometimes it converts — but it's also the most common home of junk clicks and accidental taps. For lead-gen and considered purchases, it's the first placement to check when spend looks high and results don't.

The deepest event you have enough data for. Optimising for purchases with 2 purchases a week gives Meta nothing to learn from — but staying on link clicks after you have real purchase volume aims delivery at clickers, not buyers. Graduate as data allows.

Common causes: budget too low for the audience and bid, an over-narrow audience, an optimisation event that's too rare, ads stuck in review, or a low-quality ranking suppressing delivery. Fix the constraint rather than duplicating the campaign — duplication resets learning.

The pixel reports from the visitor's browser and is increasingly blocked by ad-blockers and iOS privacy; the Conversions API (CAPI) reports from your server and isn't. Running both, deduplicated, is the current standard — pixel-only setups are optimising on an undercount.

Meta's 0–10 score of how well your CAPI events match real people. Under ~5 means Meta is failing to attribute conversions you actually got — hurting both reporting and delivery. Higher match quality (more matched identifiers, sent properly) directly improves performance.

Attribution windows: Meta counts view-through and multi-day click conversions your store's last-click view doesn't. Neither number is a lie — they're different questions. The fix is one consistent, honest measurement view across platforms, not arguing with either dashboard.

Apple's App Tracking Transparency let users opt out of tracking, so a large share of iPhone conversions stopped being visible to the pixel. Costs rose and reported performance fell industry-wide. Server-side tracking and enhanced matching are how accounts recovered the missing signal.

Judging and editing during the learning phase — every big edit resets it, so the account never stabilises. Close behind: optimising for clicks when purchase data exists, pixel-only tracking, and letting one winning creative run to fatigue with no successor ready.

In order: creative fatigue (frequency up, CTR down), tracking health (did the pixel/CAPI break, did Event Match Quality fall), audience saturation, then external factors (seasonality, competitors, landing page changes). Most sudden drops are fatigue or tracking, not the algorithm turning on you.

Pause, don't delete — history has value, and you may want to reference or reactivate. But more importantly: decide on enough data. An ad with 300 impressions hasn't failed; it hasn't been tested.

Usually yes — fatigue compounds quietly, and rising frequency means you're paying more to annoy the same people. Even evergreen winners need refreshed variants. The exception is small retargeting pools where some repetition is inherent.

You're seeing survivor bias — the Ad Library shows what they run, not what works. Don't copy blindly: use competitor ads for angle ideas, then test against your own data. What converts for their offer, price point and audience may fail for yours.