The single biggest lever on this list isn't a setting most people check first: iOS's App Tracking Transparency prompt means a real slice of your conversions are invisible to Meta unless Conversions API is actually deduplicating against the browser pixel, not just installed alongside it. Get that wrong and every other number on this page, from Advantage+ budget allocation to audience overlap, is being optimized against undercounted data. Fix that first, then work through structure, audiences, creative, and placements below, and read the last section: none of the first five control whether someone actually stops scrolling when your ad appears.
01
Campaign & budget structure
Let Advantage+ campaign budget find spend across ad sets
Manually splitting budget across ad sets forces Meta's delivery system to guess where conversions will come from instead of learning from live results. Campaign-level budget with a handful of well-differentiated ad sets consistently outperforms rigid per-ad-set caps.
Check for audience overlap before blaming rising CPMs
Two ad sets quietly competing for the same people inflate your own auction against yourself. Use Audience Overlap in Ads Manager before assuming a rising CPM means the channel is saturated. Often it just means you're bidding against your own campaign.
Give a new campaign 50 conversions before judging it
Meta's delivery system needs a real outcome signal to exit the learning phase. Pausing or restructuring before that point means every result you're reacting to is still noise, not signal.
02
Audiences & targeting
Default to Advantage+ audience over narrow manual targeting
Meta's own delivery data now consistently outperforms manual interest stacks for most advertisers. The algorithm sees signals (recent behavior, device, time of day) no manual targeting selection can replicate. Manual targeting still has a place for genuinely niche B2B audiences, but it's the exception now, not the default.
Exclude recent converters from prospecting
Without an exclusion, prospecting campaigns keep re-showing ads to people who already bought, wasting budget and quietly dragging down your real cold-audience conversion rate in the reporting.
Keep remarketing windows honest to the actual sales cycle
A 7-day remarketing window on a $500 considered purchase closes before most buyers finish deciding. Match the window to how long people actually take to come back, not a default value nobody revisited.
03
Creative testing & rotation
Run Dynamic Creative before hand-picking a winner
Dynamic Creative auto-tests combinations of images, headlines, and copy and reports which pairings actually perform. This finds non-obvious winning combinations a human wouldn't have thought to pair manually.
Refresh creative before frequency crosses roughly 3-4
Frequency climbing past 3-4 in a remarketing or narrow audience is the earliest reliable signal of creative fatigue. CTR drops and CPMs rise before anyone "feels" the ad getting old. Check frequency weekly on any always-on campaign.
Build creative for the placement, not one asset stretched everywhere
A horizontal video cropped into a Story or Reel loses the framing that made it work in Feed. Meta will auto-resize if you let it, but a purpose-built vertical cut of your best-performing creative consistently beats an auto-cropped one.
04
Placements
Start with Advantage+ (automatic) placements, not a hand-picked subset
Manually excluding placements based on a hunch ("Audience Network feels low-quality") usually just removes cheap, still-converting inventory. Let automatic placements run and only exclude a specific placement after you've confirmed it in your own data, not on reputation.
Check placement-level breakdown before blaming the whole campaign
A campaign's blended CPA can look mediocre while one placement is excellent and another is dragging it down. Break results out by placement before restructuring the whole campaign over a problem that's actually isolated to one spot.
05
Tracking & measurement
Run the Conversions API alongside the Pixel, not instead of it
Since Apple's App Tracking Transparency changes, browser-only Pixel data alone reliably undercounts conversions. Server-side Conversions API (CAPI) recovers events the Pixel misses. Running both together (Meta deduplicates automatically) gives Advantage+ delivery a materially better signal to optimize against.
Check Event Match Quality, not just whether events are firing
An event can technically fire and still carry a low match score if it's missing parameters (email, phone, external ID) Meta needs to match it to a real person. Higher match quality directly improves how well delivery can find more people like your converters.
06
Bidding & measurement discipline
Start on Highest Volume, add a cost cap only once you have a real CPA benchmark
A cost cap set before you know your actual sustainable CPA either throttles delivery to nothing or gets ignored by the auction. Run unconstrained first, establish your real number from actual spend, then cap.
Judge a bid-strategy change after a full week, not day two
Bid and budget changes reset the learning phase and destabilize delivery for several days. Reacting to day-two results after a change means judging the noisiest part of the whole cycle.
07
The step every checklist skips: the ad itself
Advantage+ delivery, audience exclusions, placement mix, CAPI: all of it optimizes distribution, making sure the right person sees your ad, cheaply, and that Meta can measure it. None of it touches persuasion: whether the hook, the offer, and the copy actually earn a stopped thumb in the first half-second of the feed.
That's a creative and behavioral-psychology question, not a delivery-settings one. It's also the fastest check on this whole list. Drop in your ad and get a scored, ranked read in under a minute.
Should I still use manual audience targeting on Facebook and Instagram in 2026?
Mostly no. Advantage+ audience now outperforms manual interest stacks for most advertisers because it sees real-time signals a manual list can't replicate. The exception is genuinely niche B2B targeting where the audience is small and well-defined enough that broad targeting would waste most of its reach.
How do I know if my creative is fatigued versus my targeting being wrong?
Check frequency first. Rising frequency (past roughly 3-4) with falling CTR on the same audience points to creative fatigue, so refresh the creative before you touch targeting. If frequency is low and performance is still weak, the problem is more likely the audience or the offer, not fatigue.
Do I need the Conversions API if my Pixel already looks like it's tracking fine?
Yes. A Pixel that appears to be firing correctly in Events Manager is still very likely undercounting since Apple's App Tracking Transparency changes. The gap doesn't show up as an obvious error, it just shows up as conversions Meta never learns from. Running CAPI alongside the Pixel is the fix, not a replacement for it.
What's a realistic amount of time to judge a new campaign?
Give it roughly 50 conversions or a full week, whichever comes later, before making structural changes. Anything judged sooner is being judged during the learning phase, when results are still noisy by design.
I've fixed all of this and performance is still flat. What's left?
Almost always the creative itself. Structure, audiences, placements, and tracking all optimize distribution: whether the right person sees your ad, cheaply and measurably. None of them touch whether the ad itself makes someone want to stop scrolling. See the section below.