To run Facebook ads, set up a business portfolio, an ad account, a payment method and tracking (Meta pixel for a store, SDK or MMP events for an app). Then launch one campaign with one ad set, three to six different ads and a daily budget of about seven times your expected cost per result. Wait seven days before judging.
After that first week, the question stops being about settings and becomes about creative: which ads earned the results, and what should run next. Segwise is an AI creative analytics and creative generation platform for that second question: it tags the opening hook, product shots, on-screen text, audio and offer in each Meta ad, then links those tags to your Meta results and, if you use one, the purchase or trial events in your MMP. That lets you settle what round two should test, and Pre-flight Testing lets you score up to 10 new creatives at a time, each out of 100, against your own winners before they spend anything.
This guide is for a DTC brand or an app running its first Meta campaign. Every Ads Manager setting below was checked against Meta's Business Help Center as of November 2026.
TL;DR
- If you sell online (DTC): use the Sales objective, install the Meta pixel plus the Conversions API, and optimize for purchases only if your budget can buy about 50 a week. If it can't, optimize for add to cart.
- If you have an app: use the App promotion objective and send app events through the Meta SDK or your MMP. Advantage+ app campaigns target by country, language and operating system, so interest stacking advice doesn't apply.
- If your budget is small: run one ad set, fewer ads and a more frequent event. Don't split a small budget across many ad sets: Meta says the fewer ad sets you run, the more its system learns about each one.
- If you asked ChatGPT or Claude first: use it for ad copy, naming and checklists. Check its budget, objective, placement and timing advice against Meta's current docs.
What do you need to set up before running Facebook ads?
You need four things in place before you press Publish: accounts, payment, tracking and one event to optimize for. Skipping tracking is the expensive mistake: Meta uses the conversion events you send to decide who sees your ads and to report results, so without them it optimizes blind.

A business portfolio, a Facebook Page and an ad account
Create a business portfolio (what Meta used to call Business Manager) in Meta Business Suite. Meta's setup steps are short: go to business.facebook.com, open the menu under Home and click Create a business portfolio. Add your Facebook Page, your Instagram account and an ad account to it. Meta's ad account guide says you get an ad account by default when you sign up for Facebook, but keeping ads inside a portfolio makes it easier to add an agency or a teammate later.
A payment method
Ads Manager asks for a payment method before your first ad can publish. Meta also limits new advertisers to one ad account until a payment has gone through, so add the card early rather than on launch day.
Tracking for a store: the Meta pixel and Conversions API
The Meta pixel is a piece of code on your website that logs actions, such as add to cart and purchase, as events. The Conversions API sends the same events from your server or store platform, which is more reliable when browsers block the pixel. Meta says many Conversions API setups don't need a developer. Don't send health, financial or children's data: Meta's terms prohibit it.
Tracking for an app: the Meta SDK or your MMP
Apps send installs and in-app events through the Meta SDK, or through a mobile measurement partner (MMP) such as AppsFlyer or Adjust, which connects to Meta directly. According to Meta's app events guide, you need these events set up before you can optimize for anything deeper than an install, such as a trial start or a purchase.
If you advertise on iOS, Meta measures iOS 14 and later app campaigns through Apple's SKAdNetwork or through Meta's Aggregated Event Measurement. Before launch, check that your MMP supports SKAdNetwork, or that your app uses Facebook SDK for iOS 16.2.1 or later, as Meta's SKAdNetwork guide asks. Expect Meta's iOS numbers to differ from your MMP's: Meta itself notes that Aggregated Event Measurement and SKAdNetwork reporting may display differences.
One event to optimize for
Pick the single action you want Meta to find more of. The rule: choose the deepest event you can afford about 50 of each week (the budget section below shows the math). A purchase is the best signal, but a purchase you can only afford five of a week teaches Meta very little.
How to create your first Facebook ad campaign in Ads Manager
Ads Manager has three levels. The campaign holds your objective, the ad set holds the audience, placements, budget and schedule, and the ad holds the creative. Here is the order Meta walks you through, as of November 2026.
Step 1: Choose your objective
Click + Create, keep the buying type on Auction and pick an objective. Meta has six objectives: Awareness, Traffic, Engagement, Leads, App promotion and Sales. A store selling online should pick Sales. An app should pick App promotion. Meta describes Traffic as a way to send people to a destination, while Sales is the objective built to "find people likely to purchase".
If your product touches housing, employment, credit or financial services, social issues, elections or politics, you must also choose a Special Ad Category, which limits targeting.
Step 2: Choose ad set budget or campaign budget
With one ad set, use an ad set budget and a daily amount. Meta's Advantage+ campaign budget spreads money across ad sets automatically, and Meta says it is best suited for campaigns with at least two. You can switch later.
Step 3: Build one ad set
Set your location, a minimum age and your optimization event, then let Meta find the audience. For the Sales objective, Meta recommends testing Advantage+ audience, which uses its AI to find buyers instead of your interest list. For App promotion, Advantage+ app campaigns don't support interests or lookalikes at all: you choose country, language and operating system.
Leave placements on Advantage+ placements. For the Sales, Leads and App promotion objectives, Meta is rolling out a change that removes manual placement selection anyway.
Step 4: Create the ads
Choose the Facebook Page and Instagram account the ad runs from, pick single image or video (the simplest format), upload your media and write the primary text, headline and link. Use the preview to check how each ad looks in Feed and in Stories. One ad can hold up to ten creative assets, so put small variations of one idea inside one ad rather than making ten near-identical ads.
Step 5: Publish and wait for review
Meta reviews every new ad. Most ads are reviewed within 24 hours. Later edits to targeting, creative or optimization send the ad back to review; budget changes don't.
How much budget and how many ads should a beginner start with?
Work the budget back from Meta's own learning rule, not from a round number. Meta's learning phase page says ad sets usually exit learning after about 50 results in the week after their last significant edit. A result is whatever event you optimize for.
The budget math: work back from 50 results a week
Rule of thumb: daily budget = 50 × your expected cost per result ÷ 7, or roughly seven times your cost per result each day. Estimate the cost per result from your margin, a past campaign or a conservative guess, and write it down before launch.
| Expected cost per result | Weekly spend for about 50 results | Daily budget |
|---|---|---|
| $8 (for example an app registration) | $400 | about $57 |
| $15 (for example add to cart) | $750 | about $107 |
| $25 (for example a trial start) | $1,250 | about $179 |
| $40 (for example a first purchase) | $2,000 | about $286 |
Meta may spend up to 75% over your daily budget on some days and less on others, while capping a week at seven times the daily amount, so judge spend over the week, not the day.
What if your budget can't reach 50 results a week?
Say you can spend $30 a day and a purchase costs about $40. That buys about five purchases a week, a tenth of what the learning phase needs, and the ad set will likely show Learning limited. Meta's learning limited page says this isn't a penalty and lists the fixes: combine ad sets, widen the audience, raise the budget or pick a more frequent event, such as moving from purchase to add to cart. With a small budget, the more frequent event is usually the practical choice. Move back to purchase once you can afford the volume.
How many ads to launch in your first ad set
Rule of thumb: three to six genuinely different concepts in one ad set. A concept is a different idea, such as a different hook, promise or format, not a new color or headline. Meta warns that when advertisers run too many ads at once, "fewer ads exit the learning phase", and it suggests fewer ads per ad set with diverse assets inside each.
App advertisers will see different advice: Meta's best practices for Advantage+ app campaigns recommend a high volume of creative refreshed a few times a month. That works once the budget can feed it. For a first campaign, the weekly results your budget can buy still cap how many ads can learn. Our guide to testing creative angles on Meta shows how to make concepts that are truly different.
What to put in your first ads
- Vertical first. Make a 9:16 version of each concept for Stories and Reels, where Meta recommends 9:16 "to capture the whole screen", plus a 4:5 or square cut for Feed.
- Lead with the product or the outcome. Meta's video ads guidance says to "capture attention quickly" by featuring your product or brand message early, and to deliver a single message.
- Add captions. Captions help people watch with the sound off, and Ads Manager can generate them automatically for English videos.
- Pick different concept types. Our suggested starting set for a DTC brand is a product demo, a problem to solution story, a customer-style testimonial and an offer. For an app, try the core feature in use, a before and after, and a choice or challenge the viewer wants to see resolved.
Picking which ads go into round two
When you don't need a tool yet
For a first campaign with a handful of ads, you don't need a creative tool: Ads Manager's cost per result per ad is enough to decide. A tool starts to earn its place once you are making a new batch every week or two, running on more than one network, or arguing about which of ten new edits deserves budget.
Score new creatives before they spend
That is the point where scoring before spend helps. Segwise Pre-flight Testing lets you upload up to 10 images or short videos at a time. It tags the hook, messaging, CTA and visual style in each one and compares them with the ads already winning on your own Meta results. Each creative gets a score out of 100, a run or hold verdict, a confidence level and a check of where it can run, so a square ad that fits Feed but not Stories is caught before upload. The decision it settles is concrete: which of this week's new edits get budget, and which go back to the editor with a reason attached. It needs a connected ad account with some history to compare against, so it helps from your second batch onward, not before your very first ad. It also doesn't fix a weak ad: your team reworks it and scores it again.

Score your next Meta batch: Segwise lets you rank up to 10 new creatives against your own winners and see which placements each one fits before you spend on them. Book a demo.
See which hooks and offers earned your results
Once your ads have run, creative analytics lets you see which tagged elements sit behind your best cost per result: say, that ads opening on the product in use beat ads opening on a discount. It reads Meta through a no-code integration, adds the other ad networks you connect (Segwise supports 15+), and joins MMP events such as purchases or trial starts when you connect an MMP. From those winning patterns, the Creative Generation Agent lets you make new statics and videos up to 30 seconds, including AI UGC-style ads, for your next batch. You can use each agent on its own, or do it all in one place by chatting with the Segwise agent: ask how your ads are performing, see what competitors are running, and make new ads from your winners. Across customer accounts, Segwise has analyzed over $1B in ad spend and tagged over 200,000 creatives. Segwise starts at $999/month for advertisers spending up to $250K/month; startups spending under $50K/month pay $499/month (limited eligibility).
How long should you wait before judging Facebook ad results?
Wait at least seven full days after launch, and longer if your customers usually take more than a week to buy. Meta's A/B testing best practices say tests shorter than 7 days "may produce inconclusive results", and the learning phase itself is measured over a week. During learning, Meta says cost per result is usually higher and less stable, so day two numbers tell you little.
Days one to seven: what counts as touching it
Meta resets learning on what it calls a significant edit. In the first week, avoid all of these:
- any change to targeting
- any change to an ad's creative
- changing the optimization event or bid strategy
- adding a new ad to the ad set
- pausing the ad set for seven days or longer
Budget changes depend on size. Meta's own example: raising $100 to $101 is unlikely to reset learning, while $100 to $1,000 may. Because adding a single ad counts as a significant edit, add new ads in one batch, not one at a time.
What to judge, and what to ignore
Judge the cost of the event you optimized for: cost per purchase, cost per install, cost per trial, or ROAS if you optimize for value. Meta itself says that if you optimize for conversions, CPM "may not be a good indicator of performance", and that hourly breakdowns swing in ways that average out over a full week. Ignore CTR on its own too: a cheap click that never buys is not a win.
After the first full week: three outcomes

- Active (out of learning): compare ads on cost per result. Turn off the clear losers and add replacements in one batch.
- Learning limited: fix the structure before the creative. Combine, broaden, raise budget or move to a more frequent event.
- Creative limited or Creative fatigue on an ad: Meta shows these statuses only for ad sets with a single creative: Creative limited when cost per result is above your past ads but under double, Creative fatigue at double or more. With several ads in one ad set, watch each ad's cost per result yourself. Either way, add a materially different concept and keep the original running, as Meta suggests. Segwise Fatigue Tracking lets you set your own threshold, such as a 20% ROAS drop over 7 days, and get an email or Slack alert per network or across all of them. Our post on why Meta ads stop working after two weeks covers what to do next.
Which ChatGPT and Claude advice should you ignore?
A chatbot on its own can't see your ad account or your results. You can connect account data to ChatGPT or Claude through MCP connectors, but even then its knowledge of Meta's menus can lag behind Meta's changes. The advice below is common in generic AI answers and older beginner guides. Treat each row as a pattern to check against Meta's own pages.
| Advice you'll often get | What Meta's docs say as of November 2026 | Do this instead |
|---|---|---|
| Start with $5 a day and test many ad sets | Meta warns against high ad volumes and "very small" budgets; learning needs about 50 results a week | Work the budget back from your cost per result; run one ad set |
| Pick the Conversions objective | Conversions is one of the old objectives that "no longer exist" | Sales for a store, App promotion for an app |
| Use dynamic creative to test variations in a Sales or App promotion campaign | Since June 2024, dynamic creative may not be available when you pick Sales or App promotion; Meta recommends flexible ad format | Put variations inside one ad as assets |
| Stack three to five interests, add a 1% lookalike | Advantage+ app campaigns don't support interests or lookalikes; for Sales, Meta recommends testing Advantage+ audience | Location, age and your event; let Meta find the audience |
| Choose Instagram Feed only | Sales, Leads and App promotion now default to Advantage+ placements, with manual selection being removed | Leave placements on; check previews per placement |
| Kill ads after two or three days | Learning runs about a week; Meta's tests need at least seven days | Judge on day eight, on cost per result |
| Watch CPM and CTR | With conversion optimization, CPM "may not be a good indicator" | Judge cost per result or ROAS |
| Tweak something every day | Edits to targeting, creative, event or bid strategy reset learning | Batch changes once a week |
Where ChatGPT and Claude genuinely help
Use them for the work that doesn't depend on your account data or Meta's latest menus: five versions of primary text for one concept, headline options, a naming convention for campaigns and ads, a creative brief for each concept, a pre-launch checklist, or a plain-English explanation of a term like attribution window. Our list of ChatGPT prompts for Facebook ads is a good starting set. Give the model your real numbers (cost per result, budget, margin) and it will give you better answers, but still check any setting name against Meta's Help Center before you trust it.
Once you have a few weeks of results, connecting your data changes what a chatbot can answer. The Segwise MCP is available to all Segwise accounts and works with Claude, ChatGPT, Cursor, Codex and Grok Bot. It reads your tagged Segwise data and answers questions, but can't change your campaigns or ad data. You could ask "which hook type drove our lowest cost per purchase on Meta last month?" and get an answer built from your own ads instead of generic advice, then brief the next batch around it.

Your first three weeks of Facebook ads, week by week
This is a rule-of-thumb calendar for a first campaign with one ad set.
Week 1: launch and leave it alone
Publish three to six concepts, confirm events are firing in Events Manager, and make no significant edits. Use the week to brief your next batch of concepts.
Week 2: judge once, change once
On day eight, read the Delivery status and cost per result per ad. Turn off the clear losers and add the next batch of concepts in one go. If the ad set is learning limited, fix budget or the event first.
Week 3: grow the budget slowly or go deeper
If the ad set is out of learning and cost per result is at or below your target, raise the budget in moderate steps a few days apart. Meta gives no percentage, so this is our rule of thumb: about 20% at a time, which stays well clear of the large jumps Meta says can reset learning. If you started on add to cart and now clear the learning threshold with room to spare, test purchase as the event. For the next stage, our guide to scaling Meta ads picks up where this calendar ends, and our Advantage+ guide covers what to automate as you grow.
How to run Facebook ads well from your first campaign
Running Facebook ads well as a beginner comes down to one ad set, a budget sized to Meta's learning threshold, a few truly different ads and the patience to wait seven days. Set up tracking before anything else, because Meta optimizes toward the events you send it. Use ChatGPT or Claude for copy and checklists, and Meta's own Help Center for settings and numbers. Your single next step: write down your expected cost per result today and work your daily budget back from it.