This is our Meta playbook, currently the only platform deep-dive. Google and TikTok guides will follow. The principles carry across channels, but the specifics here (objectives, attribution, the learning phase) are Meta’s.
These are field-tested rules of thumb, not laws. Meta is, in its own words, part science and part art, so treat them as starting points and adapt to your own results.
How Meta works now
Targeting used to be the game. You hand-picked interests and lookalikes, and the creative came second. That flipped. Meta is now creative-led: you give it broad targeting and a steady supply of creative, and the algorithm reads each ad to decide who sees it. Talk to runners in the ad, and Meta finds runners. Your job is no longer to find the audience. It’s to feed Meta enough good creative that it can find the audience for you. That is why creative strategy is the real lever. Every account has three layers, and all three have to be on for an ad to deliver:Campaign
The objective you optimize for, usually purchases or sign-ups.
Ad set
Budget, audience, and placements. Where targeting lives.
Ad
The creative people actually see.
Set your target CPA first
Two numbers drive every decision: CPA (cost per acquisition, what you pay for one customer) and LTV (what that customer is worth over time). There is no universal “good” CPA. The right target is the one your own economics support, and you work it out from three things:What a customer is worth
Your order value or LTV, minus the cost to deliver. That leaves the margin you keep.
How fast you recoup it
Your payback window. If you earn it back on the first purchase, you can afford a higher CPA than if it arrives slowly over months.
How aggressive you want to be
Targeting close to break-even buys growth. Aiming well below it protects profit per sale. Both are valid, as long as the choice is deliberate.
Attribution: read Meta’s numbers with a pinch of salt
Attribution is how Meta decides which sales to credit to your ads. The current default is 7-day click, 1-day engage-through, 1-day view. It credits a conversion when, beforehand, someone clicked through within 7 days, interacted (a like, share, save, or a few seconds of video) within 1 day, or simply saw the ad within 1 day. The thing to understand is the incentive. Every platform wants to credit itself with as many conversions as it can, because the better Meta looks, the more you spend there. If you run one or two channels at modest spend, don’t overthink it; the default is right. It only gets hard at scale, when several channels run at once and each claims the same sale. At that point you stop trusting any single platform’s self-report and reconcile against your actual total revenue, usually at month-end.The operating loop
Good media buying runs as a loop, in two lanes:Testing lane
A controlled, lower-budget place to try new concepts, hooks, formats, and audiences without contaminating winners.
Scaling lane
A higher-budget lane for proven winners and close iterations on what already works.
Monitoring cadence
A recurring check for fatigue, waste, winners, and the next action.
Learning context
A record of what worked, for whom, and what to make next.
Budgeting and the learning phase
- Budget each test at ~2–3× your target CPA per concept, per day. That buys enough conversions to get a clear read in days rather than weeks.
- Mind the learning phase. After a launch or a significant change, Meta re-optimizes for two to three days, and CPA swings around before it settles. Don’t judge or react during that window. Meta officially treats an ad set as out of learning once it gets about 50 optimization events in 7 days. Smaller budgets often never hit that and run in “limited learning,” which is normal; just read thin numbers more cautiously.
- Scale in small steps. Raise a winner’s budget by up to about 10% a day, and never by more than 19%. A bigger jump counts as a significant change and re-triggers learning. Wait a full 24 hours between increases. (Those percentages are rules of thumb, not limits Meta publishes.)
- Split roughly 30% testing, 70% scaling, and move weight back to testing when scaling stops outperforming it.
Reading results
- Ignore learning-phase noise for the first few days, and after any significant change.
- Watch frequency, and keep it below ~5 (healthy is around 1.5–3). Rising frequency is your cue to ship fresh creative, not to push more budget.
- When CPA runs hot past the learning phase, the fix is usually new creative, not more spend on a tiring winner.
Ask for action, not a dissertation
A useful report tells you what to do today:- Pause these ads.
- Scale these winners.
- Create these follow-up variants.
- Test this new hook.
- Watch this audience or placement.
Human approval still matters
Use Superscale to compress the work: research, creative generation, analysis, and recommendations. Keep humans in the loop for budget, brand risk, compliance, and final approval.How Superscale helps: it keeps the top of this loop full, researching winning angles, generating concept-and-variation batches, publishing to Meta, and surfacing what to pause or scale. Spend-changing actions stay approval-gated, so you sign off before anything goes live.
Related
Campaign structure
The two-lane setup, budgets, and reuse matrix.
Creative strategy
How to generate the concepts this all runs on.
Performance monitoring
Track winners, waste, and fatigue.
Publishing
How launch and approval work.