A Facebook ad campaign is built on three nested levels — campaign, ad set, ad — and each owns a different decision. The campaign level sets the objective and, optionally, a shared budget; the ad set level defines who sees the ads, on which placements, over what schedule, at what budget, and which action Meta optimizes toward; the ad level holds the creative — image or video, primary text, headline, call-to-action button, destination URL. Build in that order: the objective determines which optimization events appear in the ad set below it, and the optimization event determines who the delivery system shows your ads to.
The interface work is the last ten minutes of the job. Objective, audience, message, budget and the metric that settles the verdict belong in a document written before the account opens. The Advertising Major's Campaign Workbook calls the alternative the execution-first trap: designing posts before establishing strategic foundations, which yields polished work with no coherence.
What each of the three levels decides
Most beginner campaigns go wrong here, one altitude above the ad.
| Level | What you decide here | What you must NOT decide here | Typical mistake |
|---|---|---|---|
| Campaign | Objective, special ad category, whether budget is shared across ad sets, A/B test on or off | Who the ads reach, what they say, how much any one audience gets | An objective matching the ad's format (video views) instead of the outcome (purchases) |
| Ad set | Audience, placements, schedule, budget if not set above, optimization event, attribution, bid strategy | The message, the offer, the angle — and the objective, locked above | Splitting one audience into six narrow ad sets, so none gets enough conversions |
| Ad | Format, image or video, primary text, headline, description, call-to-action button, destination URL, tracking parameters | Who sees it, what it costs, when it runs | Running one creative, leaving nothing to compare and no way to diagnose |
Settle the brief before you open Ads Manager
The workbook's creative brief has five parts, four of which map onto decisions the platform forces on you anyway.
- Background. Why this campaign exists now and what has already been tried — the part that stops you re-running a message that failed.
- The objective, as a number with a deadline. The workbook's benchmark: "Increase website trial signups by 25% among first-time visitors from paid social channels within Q2." Compare "raise brand awareness," which can neither guide a creative decision nor settle whether the campaign worked.
- The audience, defined behaviorally. "People who have considered switching from Brand X but haven't acted because switching feels overwhelming" is actionable. "Adults 25–40" is a demographic filter — and the filter is all the Ads Manager panel asks for. You translate the behavioral definition into interests, custom audiences or a deliberately broad setting; the interface never prompts you to write it.
- The single-minded proposition. One message every ad approaches from a different angle — "Switching is easier than you think" is a sentence a copywriter can execute from. Campaigns carrying several messages dilute all of them.
- Mandatories and guardrails. Disclaimers, logo usage, tone limits — plus one hard Meta constraint: offers touching credit, employment, housing, or social issues, elections and politics must be declared a special ad category, which restricts age, gender and location targeting and disables lookalikes.
Then the brief approval test: hand it to someone unfamiliar with the project and ask them to state the objective, audience and core message after one read. If they cannot, it cannot guide a media buy either.
The objective menu, and how the wrong one wastes budget
Meta consolidated its objectives into six outcome-based options: Awareness, Traffic, Engagement, Leads, App promotion and Sales. Delivery optimizes toward the event you name, and it is very good at finding people who perform that specific event — including people who perform it and nothing else. Choose Traffic and you buy cheap clickers — people with a habit of tapping links — producing a low CPC and often a terrible cost per purchase, because click-likelihood and buy-likelihood are different behaviors. Choose Sales with a purchase event and the system aims at what you care about, but only if the pixel or Conversions API reports it reliably and often enough to learn from.
The objective is how you apply the workbook's channel role assignment principle inside one platform. Awareness, consideration and conversion roles belong in separate campaigns with separate objectives and success metrics, never in one campaign judged against contradictory standards.
Audience: core, custom, lookalike, and when broad wins
The workbook prescribes concentric layers: the core target gets the highest frequency, lookalikes extend reach to similar prospects, retargeting recaptures people who engaged without converting. Meta's three audience types map onto them exactly.
- Core audiences come from location, age, gender, detailed targeting (interests, behaviors, demographics) and connections — the only layer available before you hold first-party data.
- Custom audiences come from data you own: pixel-tracked visitors, uploaded customer lists, video viewers, Page or Instagram engagers. Retargeting lives here and is nearly always the cheapest conversion source in the account — which is why it must not be judged as prospecting. It harvests demand that upper-funnel spend created.
- Lookalike audiences are modeled from a custom audience seed. A 1% lookalike is the most similar slice of a country's population; 10% is broader and more diluted. The seed sets the ceiling: 200 low-value leads model low-value leads.
Narrow is usually worse. Stacking six interests, three behaviors and a tight age band feels like precision, but it shrinks the pool the system can search and prices you into a smaller auction against everyone who picked the same obvious interests. Narrow earns its keep twice: a genuinely small, identifiable audience, and retargeting a defined list. Everywhere else, let the creative target — an ad opening "If you still process invoices by hand…" self-selects more sharply than any checkbox.
Budget, campaign budget optimization, and the learning phase
Budget sits at campaign level (campaign budget optimization, now branded Advantage campaign budget) or at ad set level. The first shifts spend toward whichever ad set delivers most cheaply; the second guarantees each audience a fixed amount. Use campaign level when ad sets are interchangeable, ad-set budgets to stop a cheap retargeting audience from starving your prospecting.
The constraint governing the rest is the learning phase. A new or heavily edited ad set delivers unstably while the system explores; Meta's documented threshold for exiting is roughly 50 optimization events per ad set within seven days. Below it the ad set stays "learning limited" and cost per result is usually worse.
- Minimum viable budget follows from cost per event, not from comfort. At a $30 purchase cost, 50 purchases a week costs $1,500 — about $214 a day — for one ad set. If that is out of reach, do not spread the shortfall across four ad sets; run one, or optimize for a cheaper upstream event (add to cart, lead, landing page view) that clears 50 a week at your real budget.
- Every significant edit restarts learning. Changing the optimization event, audience, creative set, or budget by a large margin resets the phase. Edits are never free.
The workbook's 70/20/10 model splits the budget sensibly: 70% to what reliably works, 20% to formats with evidence behind them, 10% to genuine experiments.
The ad, and how to test creative so the result means something
A feed ad competes with posts from the viewer's friends and has roughly the attention window of a billboard passed at speed. Import the workbook's seven-word OOH test: if the idea needs more than seven words, it is not simple enough to scale. Build each ad from three components and nothing else.
- Hook. First frame or first line, carrying the tension from the brief. The workbook's ideation methods produce hooks directly: reframing ("Your current provider is hoping you never read this"), tension amplification, metaphor ("Your data is like a house with no locks"), role reversal.
- Proof. One concrete reason to believe: a number, a demonstration, a customer's own words. Vague claims cost the same attention as specific ones and convert worse.
- One call to action, stated once, matched to the button and to what the landing page delivers. An ad offering a demo, a trial and a newsletter is three ads badly stacked.
Before production, run the workbook's brand connection test — show the concept with the brand removed and ask what brand it is for; if nobody places it in the right category, the idea is entertaining and strategically disconnected — and its rough execution test, an unpolished mockup shown to five target-audience members so production value is not doing the persuading.
The workbook prescribes three to four creative variations per ad set, with budget reallocated to winners inside 48 hours. That tempo is right, but "which ad got more sales" and "which ad is better" are different questions, and the gap is sample size. A usable test changes one variable; a new hook, image and landing page at once teaches nothing reusable. It accumulates enough conversions per arm: at twenty conversions per variant, a 40% gap between two identical ads occurs routinely by chance. And its stopping rule is written first — "run until each ad has 100 link clicks or seven days pass, then keep the higher landing-page-view rate" — because deciding when to stop after seeing the numbers is how a fluctuation becomes a belief. Meta's A/B test tool splits the audience so variants do not bid against each other; comparison inside one ad set is directional only.
Reading the metrics that matter
Assume 30 days, $1,500 at $50 a day, one ad set, purchase optimization, average order value $80. Every rate below is a stated assumption; substitute your own.
- CPM $12 → 125,000 impressions.
- Reach 45,000 people → frequency = 125,000 ÷ 45,000 = 2.8.
- Link CTR 1.2% → 1,500 clicks; CPC = $1.00.
- Landing page conversion 3% → 45 purchases; CPA = $33.33.
- Revenue = 45 × $80 = $3,600; ROAS = 2.4.
The plan fails before it runs: 45 purchases is about 11 a week, far under the 50-per-week threshold, so this ad set lives its whole life in learning limited. The fix is structural — consolidate, optimize for a cheaper mid-funnel event, or raise budget. And ROAS 2.4 is good news only if gross margin exceeds roughly 42%, since break-even ROAS is 1 ÷ margin.
| Metric | What it measures | What a bad number tells you | First thing to fix |
|---|---|---|---|
| CPM | Cost per 1,000 impressions: the price of attention in your auction | Rising CPM with unchanged creative means auction pressure, or too narrow an audience to bid into cheaply | Broaden the audience; unstack targeting |
| CTR (link) | Share of impressions producing a click to your destination | Low CTR is creative, not targeting: the ad reached people and failed to interest them | Rewrite the hook and first frame; test another angle |
| CPC | What a visit costs — effectively CPM divided by CTR | High CPC with decent CTR points back at CPM; with low CTR it points at creative | Diagnose CPM and CTR separately before touching bids |
| CPA / CAC | Cost to acquire one conversion or customer | Good CTR with bad CPA means the ad sold what the page does not deliver | Fix message match; check offer, price, form length |
| ROAS | Revenue divided by ad spend | ROAS under break-even means unit economics, not the campaign, are the problem | Compute break-even from margin, then raise order value or cut CPA |
| Frequency | Impressions ÷ reach — how often the average person saw the ad | Rising frequency with falling CTR is creative fatigue in too small a pool | Add creative or expand the audience; more budget alone speeds decay |
The workbook's AECR framework ties metrics to objectives — Awareness (reach, video views, branded search lift), Engagement (engagement and share rate, saves, sentiment), Conversion (conversion rate, CPA, revenue), ROI (ROAS, acquisition cost against lifetime value) — and judging an awareness campaign by CPA is the commonest measurement error in paid social. Its companion is the pre-campaign baseline: record every KPI before launch, or you report absolute numbers but never demonstrate lift.
The mistakes that cost beginners the most
- Set-and-forget — or its opposite. The workbook's cadence: a 24-hour technical check (ads serving, pixels firing, right audience reached), 48-hour creative optimization, weekly channel review. Editing more often is the mirror-image error, since every change restarts learning.
- Fragmenting the account. Eight ad sets at $10 a day is eight ad sets permanently in learning. Consolidation is usually the highest-leverage fix a small advertiser has.
- Optimizing for the easy event. Link clicks are cheap, plentiful, and the metric most detached from revenue.
- Skipping tracking hygiene. The workbook's unified tracking framework — consistent UTM parameters, conversion pixels and attribution across channels — is what makes comparison possible. On Meta: verify your domain, configure your eight prioritized web events under Aggregated Event Measurement, and know your attribution setting (default 7-day click, 1-day view) before trusting platform conversion counts. The workbook's channel bias trap applies here too: allocate to where the data is, not to the formats you enjoy.
Frequently asked questions
What is the minimum daily budget for a Facebook campaign?
Meta enforces a technical minimum that scales with what you optimize for, but that is not the useful number. Multiply your expected cost per event by 50 and divide by seven: that is the budget giving one ad set a fair chance of leaving the learning phase, roughly $214 a day at $30 per purchase. If it is impossible, optimize for a cheaper event rather than starving several ad sets at once.
How long should I run a campaign before judging it?
Check delivery at 24 hours, decide on creative at 48, and reserve the verdict until the ad set has exited learning and accumulated enough conversions to detect the difference you care about — one to two weeks on most small budgets. Judging on day three judges the exploration period, the noisiest data the campaign produces.
How many ads should I put in one ad set?
Three to four: enough for the system to find a winner, few enough that each gets meaningful delivery. Beyond that, budget spreads too thin to evaluate most variants. Vary one dimension at a time so the win transfers.
Why do my costs rise after the first week?
Usually frequency. As the ad set exhausts the most responsive people in a finite audience, the same viewers see the ad repeatedly, CTR falls, the estimated action rate falls with it, and you pay more for the same delivery. That combination is creative fatigue: the fix is new creative or a larger audience, not a higher bid. Seasonal auction pressure raises CPM independently.
Do I need a Business Manager account?
You need a Facebook Page and an ad account. A personal account works for a solo advertiser, but a Business Manager (reached through Meta Business Suite) lets you own assets at business level, grant and revoke access by role, verify your domain, and manage the pixel and Conversions API centrally. Domain verification is required to configure prioritized web events, so conversion campaigns need it. Boosting a post needs none of this and gives no ad set structure, which is why it suits amplifying an organic winner and nothing more.
All of this is the paid-media slice of a larger discipline: the brief that defines the message, the ideation methods that generate concepts worth testing, the media mix that decides how much budget Meta should receive at all, and the analysis that turns one result into transferable learnings. The Advertising Major's Campaign Workbook is the complete guide to that arc — the five-part brief, the SESM test, the concept-testing methods, the 70/20/10 model, the AECR framework, and a 60-day sprint from brand selection to pitch deck. This article helps you set a campaign up correctly; that book is where you work out what it should have been.
