How to Set Up a Facebook Ad Campaign (2026)

Campaign, ad set, ad: which decision lives where

Every campaign you build in Ads Manager is the same three-level hierarchy, and most setup mistakes are really a decision sitting at the wrong level. The shorthand worth memorizing: the campaign decides what you are buying, the ad set decides who sees it and for how much, and the ad decides what they see. That separation tells you where every future change belongs: a new audience means duplicating an ad set, a new hook means adding an ad, and a different outcome means a new campaign, because the objective is fixed once a campaign is live.

Before you open Ads Manager: account and compliance

A campaign assumes a few things already exist: a Meta Business account holding the assets and billing, an ad account inside it, and a Facebook Page (ideally a linked Instagram account) to run from. One ad account is plenty to start, and the ceilings are generous: a standard account allows up to 5,000 campaigns, 5,000 ad sets, and 5,000 active ads, 50 ads per ad set (Armada Growth, 2025). What bites first is your budget, not the account.

One setup screen can quietly get an account restricted: the Special Ad Category declaration. If your ad relates to credit or finance, employment, housing, or social issues, elections, and politics, Meta requires you to declare it, which strips out age, gender, and tight location targeting (Take Flyte, 2025). The trap runs both ways: sell a financial product or job listing without declaring and the account risks suspension; tick the box for a skincare ad and you have handicapped your targeting for nothing. For most ecommerce stores the answer is "none", but read the four categories once and answer honestly.

The six objectives, and how to pick one

Since 2022 Meta has run on six objectives, the Outcome-Driven Ad Experiences (ODAX) set, consolidated from the 11 legacy options older guides still describe (Social Media Examiner, 2025). The old labels did not vanish; they collapsed into the six as optimization choices and conversion locations underneath. So Conversions and Catalog Sales both live inside Sales now, Video Views and Messages live inside Engagement, and you pick the outcome first, then the specific event below it.

The objective is not a label, it is the instruction Meta optimizes toward, and each one exposes a different set of optimization events and conversion locations. Reading the table below as a menu, rather than picking the first thing that sounds right, is how you avoid paying for the wrong action.

For an ecommerce store the honest choice is Sales, optimized for purchases, though each result costs far more than a click. You are paying for the one event that pays you back. The numbers make the trap in the cheaper objectives visible: a Facebook traffic click averaged $0.70 in 2025 (WordStream). Cheap clicks look great in a report, and they are cheap because clicking is easy. A thousand $0.70 clicks that never reach checkout cost more than a handful of purchases, while training the account on the wrong people. Choose Leads when the sale closes off-platform through a conversation, Awareness when reach itself is the point, and Sales the rest of the time.

Advantage+ Sales or a manual campaign

Once you choose Sales, Ads Manager asks a second, bigger question that older guides miss: Advantage+ Sales or a manual campaign. These are not two checkboxes inside one flow, they are two different campaign types with real trade-offs. Meta renamed its Advantage+ Shopping product to Advantage+ Sales in early 2025 (Social Media Today, 2025), and the automation behind it performs because it searches further than any hand-built setup can. Meta reported its Advantage+ products crossing a $20 billion annual run rate, up 70% year over year (AdExchanger, 2025), and Facebook ads could reach 2.28 billion people as of January 2025 (DataReportal). A manual interest stack covers a sliver of that.

What you gain with Advantage+ Sales is reach and hands-off optimization. What you give up is control. In an Advantage+ Sales campaign you cannot set specific age, gender, or interest targeting; you cannot select or exclude individual placements; and geography is broadly country-level. Your audience inputs become "Audience suggestions", treated as a hint Meta may overshoot, not a fence it stays inside. A manual Sales campaign keeps all of that in your hands: the precise audience, placement selection and exclusion, and the customer-list exclusions a prospecting budget needs so it stops paying to reach people who already bought.

The decision is rarely all or nothing. Pick Advantage+ Sales for broad prospecting, for scaling a proven offer, and when a small team has no time to hand-tune. Pick manual when you sell to a niche the automation cannot infer, when a hard age or geographic constraint applies, or when ring-fencing existing customers and controlling placements is non-negotiable. Plenty of accounts run one of each in parallel.

Inside the ad set: audience, placements, optimisation, attribution

The ad set is where most of the campaign actually gets built, and the screen reads top to bottom as one connected instruction. Fill it in that order and the dependencies make sense.

Conversion location and event. First you tell Meta where the conversion happens (Website for most stores) and which event counts (Purchase for a store that already gets sales). This is the most consequential setting in the build. Without a pixel firing real purchase events, a Sales campaign is inert: Meta optimizes blind, and your reports cannot separate a winning ad from a lucky one (Shopify, 2026). Install it and confirm the event registers in Events Manager before you spend. A brand new pixel with no conversion history starts learning from scratch, so if purchases are too rare to feed the system, optimize for a more frequent upstream event (Add to Cart or Initiate Checkout) while volume builds, then graduate to Purchase.

Audience. If you went manual, this is where you build it: custom audiences (a list you own, your customer file or pixel visitors), lookalikes (Meta finds new people resembling a seed, usually a 1% lookalike of recent purchasers, widening to 3% to 5% as you scale), and the control that quietly saves the most, exclusion. A prospecting ad set pays to find new buyers, so exclude your customer list and active retargeting audiences, or the two ad sets bid against each other for the same person.

Placements. Below the audience sits placements: Advantage+ placements (the default) runs the ad across Facebook, Instagram, Messenger, and the Audience Network and shifts spend toward whatever delivers. Manual placements lets you pick and exclude each one. Keep the default unless you have a real reason, a brand-safety rule or a creative built only for vertical feeds, since narrowing placements shrinks the auction.

Attribution and optimisation. The optimization event you set above is what delivery chases. The attribution window decides which conversions get credited: the default 7-day click plus 1-day view credits a sale to an ad clicked up to a week earlier. Hold that wider window while volume is thin, tighten to 1-day click once daily sales are dense, and keep it consistent when you compare campaigns, since it changes only which sales the report credits, not what happened.

The rule for taking any control back from Advantage+ is one sentence: you must be able to finish "manual, because". A hard geographic or age constraint the automation cannot infer finishes it; a retargeting message that would make no sense to a cold audience finishes it. "Because I like control" does not.

Budget and bidding: where to put the money and how to bid it

Three settings decide how spend flows: budget at the campaign or ad set level, daily or lifetime, and which bid strategy governs the auction. All three turn on the learning phase.

One budget or many, and how wide to build

Advantage+ campaign budget (still widely called CBO) puts one budget at the campaign level and lets Meta distribute spend in real time toward whichever ad set performs; with ad set budgets, you fix the split yourself. The campaign budget is the better default: your ad sets are usually peers chasing the same outcome, and a fixed split just preserves your guess after the data has outvoted it. Ad set budgets earn their place when spend must be guaranteed: a structured test where each audience needs even delivery, or a small retargeting ad set a big prospecting one would starve.

The common objection to CBO is "I lose control of the split", and Meta answers it directly: under a campaign budget you can set a minimum or maximum spend per ad set, so a small but valuable ad set (say retargeting) is not starved while a broad prospecting one soaks up everything (Meta for Business, 2026). Two eligibility rules come with the campaign budget: every ad set must share the same budget type and the same bid strategy, on standard delivery (Meta for Business, 2026).

The budget question and the how-many question are the same. Every ad set runs its own count toward those roughly 50 weekly events Meta has long advised (Search Engine Land, 2025), so consolidation is not a style preference, it is arithmetic. Picture a budget that buys about 50 purchases a week in one ad set at a $40 cost per acquisition. Split that same spend across ten lookalike and interest ad sets and each buys roughly five purchases a week: none clears that threshold, so all ten sit in Learning Limited at once, and you have paid for ten unstable ad sets instead of one stable one. Pool the spend back into one campaign with one or two broad ad sets and the conversions pool with it. Day one for most stores is one campaign, campaign budget on, one or two broad ad sets, three to five genuinely different ads apiece.

Daily or lifetime, and the number to actually type

Above the budget box sits a second toggle: daily or lifetime. A daily budget spends roughly that amount each day until you stop it, averaging over the week. A lifetime budget sets one total across a fixed start and end date for Meta to distribute, and it unlocks ad scheduling, the dayparting control (chosen days and hours only) a daily budget lacks, so pick it for a time-boxed promotion. For an always-on store, a daily budget is the simpler default.

Then the question every guide dodges: what number goes in the box. Anchor it to the learning phase. To buy roughly 50 conversions a week, a workable floor is your target cost per acquisition times 50, divided by 7, close to seven times that cost (Top Growth Marketing, 2025). Most ad sets clear the phase inside that 7-day window, though thin volume can stretch it to two weeks or more (Modern Marketing Institute, 2026). Budget below the floor and the ad set may sit in Learning Limited. The arithmetic works in any currency; here it is in dollars.

These assume one prospecting ad set carrying the full signal; split the budget across several and each runs its own slower count.

Bid strategy: what to pick from that dropdown

A campaign budget forces every ad set to share one bid strategy, so this is a structure decision. Highest Volume (Highest Value when you optimize for purchase value) is the default and the right first choice: it spends your budget to win as many conversions, or as much value, as the auction allows, with no cost ceiling to choke delivery. Cost per Result Goal lets you name a target average cost and asks Meta to stay near it while still chasing volume; ROAS Goal does the same against a return target; Bid Cap is the manual extreme, a hard ceiling on each auction bid for margin-sensitive hands. The costliest mistake is reaching for a goal or cap on day one: set the target too tight and the system, told it may not pay above it, stops spending and never escapes learning. Launch on Highest Volume, learn your real cost, then add a guardrail.

Naming that keeps Ads Manager reporting readable

Naming feels like trivia until week six, when "New campaign (copy) (copy)" sits above "Test 2 FINAL" in a results export and nobody remembers which one held the broad audience. The fix is a convention where every level answers its own question. Campaigns carry the objective, product line, structure, and a date stamp (Sales | Skincare | CBO | 2026-06). Ad sets carry audience and geography (Broad | US | Advantage+, or Retargeting | 30d site visitors). Ads carry format and angle (Video | UGC unboxing | v3, or Static | price anchor | v1). Two habits make it stick: decide the pattern before launch, and name the idea, not the file (testimonial-mom-skin-v2 says something, FINAL_export_4 says nothing). The test is whether a stranger to your account can read one row and know what ran, to whom, and which idea it tested. Renaming never resets the learning phase, so a tidy convention costs you nothing.

The top-down build, click by click

Every decision above falls into one click order, campaign down to ad. This is the sequence Ads Manager actually walks you through:

1. Create a campaign and pick the objective. Sales for a store; the choice sets everything downstream.
2. Declare the Special Ad Category. "None" for most ecommerce; declare honestly if credit, employment, housing, or social or political issues apply.
3. Choose Advantage+ Sales or manual. Advantage+ for broad prospecting and scaling; manual when you need exclusions, niche targeting, or placement control.
4. Name the campaign. Objective, product, structure, date (Sales | Skincare | CBO | 2026-06).
5. Set the campaign budget and bid strategy. Campaign budget on, daily, Highest Volume to start. Add ad set spend minimums or maximums if one ad set must not be starved.
6. Drop into the ad set. Conversion location and event (Website, Purchase), audience, placements (Advantage+ unless you have a reason), attribution window, and the optimization event.
7. Build three to five ads. Genuinely different arguments, not crops of one photo.

That order keeps you from a common reversal: people set the budget last, after the ad set, then discover the floor they need was set by the learning phase math all along.

A worked setup: a skincare store at $12 a sale

Every decision above, in one build: a store selling a moisturizer at a $12 target cost per purchase, off a working pixel.

Four distinct creatives means four different arguments, not four crops of one photo, and skincare gives each angle a clear shape. A demo shows the product absorbing into skin, the routine a viewer would actually follow. A testimonial puts a real customer on camera describing the change, carrying the trust a brand claim cannot. A before-and-after leads with the visible result, the proof a skincare buyer scans for first. A price anchor states the offer plainly, a bundle or a launch discount, for the shopper who only needs a reason to act. Those four cover the table.

Launch it and leave it alone through the first volatile week. Once the ad set clears learning and a creative pulls ahead, scale that winner in budget steps of no more than about 20% so the change does not reset learning (AdMetrics, 2025), retire the weakest ad, and add retargeting when prospecting is healthy.

Where structures go wrong

Most underperforming accounts are not beaten by the creative, they are beaten by the structure around it. The mistakes repeat, and each has a one-line fix.

• Too many ad sets splitting the budget. The fragmentation math above is the cause: every ad set you add divides the conversion signal, and below the learning floor none stabilizes. Fix: consolidate into one or two broad ad sets and let the budget pool.
• Audience overlap from too-similar ad sets. Run a 1% lookalike beside a 2% lookalike beside a broad interest set and you bid against yourself in the same auction, inflating your own cost. Fix: keep ad sets distinct, and check the audience overlap before launch.
• Prospecting that does not exclude buyers. A cold ad set with no exclusions keeps paying to reach existing customers and your active retargeting pool. Fix: exclude the customer list and warm audiences from every prospecting ad set.
• Editing a live ad set into a learning reset. Significant edits, the audience, the optimization event, placements, the creative, or a budget jump over about 20%, send the ad set back into learning (Search Engine Land, 2025). Fix: batch necessary changes into one edit, and never tweak during the volatile first week.
• Duplicating a winner instead of scaling it. Cloning a proven ad set spawns a fresh learning phase and competes with the original. Fix: raise the original's budget in ~20% steps.
• A cost cap or bid cap on day one. Set a ceiling before you know your real cost and Meta, forbidden to bid past it, simply stops spending. Fix: start on Highest Volume, learn the number, then add a guardrail.
• Judging performance mid-learning. The first days are deliberately erratic. Fix: wait until the ad set clears the phase.

Settle the structure, then go build

Run the decisions in order and the whole structure layer closes in an afternoon. None of these choices will be the reason the campaign wins; they are the container, what fills it decides. Once the structure stands, every hour you might have spent rearranging it pays better as an hour making the next ad: a sharper hook, a different format, a fresh angle to replace whatever fatigues. A platform like AdPlay.ai keeps that production loop (research what is running, generate on-brand creative, edit it, launch to Meta) in one place, but the cadence matters more than the tool. If you have not taken an ad live before, the click-by-click walkthrough is in how to run a Facebook ad. The structure is settled. Go make the ads.