AI advertising operations

Zero to live in 72 hours: the ad launch checklist

Start with three inputs: your website URL, a spend limit and a one-line goal that names one measurable event. Hours 0-24 set up the business portfolio, ad account, payment, spending limit, Pixel, standard events and Conversions API, checked in Test Events. Hours 8-48 cover research and the first 10-20 ads across four angles and four formats. Hours 48-72 build one consolidated campaign, QA every link and leave time for ad review. After launch, day 1 asks whether it works, day 3 whether anything is broken, and day 7 what you learned.

Most first campaigns don't fail because the ads are bad. They fail because nobody can tell whether the ads are bad. The Pixel fires twice, the purchase event has no value, the budget is split ten ways, or the owner edits everything on day two.

This checklist gets you from three inputs to a campaign that spends and measures, in 72 hours:

  1. Your website URL. The page people land on after they click.
  2. A spend limit. The most you are willing to spend in the first month, in total.
  3. A one-line goal. One sentence that names one measurable action and a target cost, for example "Sell gift sets online at $35 or less per order."

The plan is written for Meta (Facebook and Instagram) because that is where most small businesses start. Every setup step links to Meta's own help center or developer documentation, checked on 16 November 2026. Meta renames things often. If a menu in your account looks different, the linked page is the current source of truth.

The workbook has the full checklist with an owner, a status and a due hour for every task. It also calculates a launch-readiness score and holds the day 1, day 3 and day 7 check sheets.

The worked example used in this guide

Every number below belongs to one made-up business, so you can follow the arithmetic. These are example values, not benchmarks or results.

InputExample value
BusinessAn online candle shop
Website URLThe shop's gift-set page
Spend limit$1,500 for the first 30 days
One-line goal"Sell gift sets online at $35 or less per order"
Average order value (AOV)$60
Daily budget$1,500 ÷ 30 = $50 a day
Measurable eventPurchase, with value and currency

The goal sentence does real work. "Get more sales" can't be measured. "Purchases at $35 or less" tells you which event to track (Purchase), which number to watch (cost per purchase, often called CPA, cost per acquisition) and what counts as success. If you sell services, use a qualified lead instead: "Booked consultations at $50 or less each."

Check the target makes sense before you start. If one order is worth $60 and the product and shipping cost $30, a $35 cost per purchase loses $5 on the first order. That may be fine if customers come back, but decide it on purpose. The allowable CAC calculator works this out for you.

The 72 hours at a glance

Gantt-style timeline of the 72-hour example plan: business portfolio, ad account and payment in hours 0 to 6; domain verification running from hour 2 to 24; Pixel and standard events in hours 4 to 12; Conversions API and Test Events in hours 8 to 20; research in hours 8 to 24; making the first 10 to 20 ads in hours 24 to 48; building the campaign and QA in hours 48 to 60; an ad review buffer in hours 60 to 68; launch and first delivery readback in hours 68 to 72

Example schedule. Waiting time, like DNS changes and ad review, runs alongside other work instead of blocking it.

The trick is to overlap the work. Domain verification can take hours because DNS (domain name system) changes need time to spread. Ad review takes time you don't control. So you start those early and keep working while they finish.

BlockHoursWhat you finishWho usually does it
Accounts0-6Business portfolio, ad account, Page and Instagram, payment, spending limitOwner
Tracking4-20Pixel, standard events, Conversions API, Test Events checkOwner or developer
Research8-24Offer sheet, audience notes, 10-20 competitor ads loggedOwner or marketer
Creative24-4810-20 finished ads in the right sizes, copy, links with UTMsMarketer, designer or AI tool
Build and QA48-60One campaign, ads attached, every link testedMarketer
Review and launch60-72Ads approved, campaign live, first delivery checkedMarketer

What has to exist before you can launch

Dependency diagram: the business portfolio comes first and leads to the ad account, the Facebook Page and Instagram account, the Meta Pixel and domain verification. The ad account leads to payment and spending limit, which leads to launch. The Page leads to the 10 to 20 ads, which lead to launch. The Pixel leads to standard events, then the Conversions API, then Test Events, which leads to launch. Domain verification sits off the launch path.

Arrows mean "must exist before". Domain verification matters, but it should never hold up a launch.

Read the diagram from left to right. Everything hangs off the business portfolio. Three chains then have to finish before launch: money (ad account, payment, spending limit), measurement (Pixel, events, Conversions API, Test Events) and creative (Page, ads). If one chain is late, the launch is late. That's why the plan works on all three at once.

Hours 0-6: accounts and money

Do this in one sitting, signed in as the person who owns the business. Turn on two-factor authentication for every admin before you start; a locked account on hour 5 costs you a day.

  • Create a business portfolio in Meta Business Suite. This is what Meta used to call Business Manager. It holds your ad accounts, Pages, Pixels and the people who can use them. Meta's guide: How to create a business portfolio in Meta Business Suite.
  • Connect your Facebook Page and Instagram account to the portfolio. Ads run from a Page, so the Page must exist and belong to the business, not to a former employee or agency.
  • Create or add the ad account inside the portfolio. Set the time zone and currency carefully; both are hard to change later. Meta's guide: Add an ad account to your business portfolio.
  • Add a second admin you trust. One admin is a single point of failure.
  • Add a payment method to the ad account. Meta's guide: Add a payment method to a Meta ad account.
  • Set the account spending limit to your spend limit. In the example that is $1,500. An account spending limit caps what the ad account can spend in total across all campaigns. When it is reached, ads stop until you raise or reset it. It is your hard safety net, separate from the daily budget. Meta's guide: Change your ad account spending limit.
  • Start domain verification now, so the DNS change has time to spread. Meta's developer documentation describes two methods, a DNS TXT record or an HTML file in your site's root folder, and explains that verification lets you control who can edit link previews that point to your domain. See About domain verification and Domain verification for developers.

Check: before hour 6, open the ad account and confirm you can see the payment method, the spending limit and two admins.

If your ads are about credit, employment, housing, or social issues, elections or politics, Meta requires a special ad category with extra rules and restricted targeting. Read How to choose a special ad category before you build anything. This 72-hour plan assumes you don't need one.

Hours 4-20: measurement you can trust

This is the part people skip, and it's the part that decides whether you learn anything in week one. The goal is simple: when someone buys, Meta should receive exactly one Purchase event with the right value.

Step 1: install the Meta Pixel and your standard events (hours 4-12)

The Meta Pixel is a snippet of code on your website that tells Meta what visitors do. Events Manager may call it a dataset; it's the same thing. Meta's guide: How to set up a Meta Pixel and events on your website.

If you use Shopify, WooCommerce or a similar store builder, use its official Meta connection instead of pasting code. If you can't edit code, Meta's event setup tool adds events by clicking on your buttons, once the Pixel base code is on the site.

Map your goal to Meta's standard event names. These names come from the Meta Pixel reference, which also says Purchase requires a value and a currency.

Your businessMain event (the goal)Supporting eventsMust include
Online storePurchaseViewContent, AddToCart, InitiateCheckoutvalue and currency on Purchase
Service businessLead or ScheduleContact, ViewContentthe lead form or booking that counts
Subscription or appSubscribe or StartTrialCompleteRegistrationvalue if you know it

Pick one main event and optimize for it. Supporting events help you see where people drop off, but your campaign should chase the one thing in your goal sentence.

Step 2: add the Conversions API (hours 8-20)

Browsers block or lose some Pixel events. The Conversions API (CAPI) sends the same events from your server, so Meta still hears about a purchase when the browser doesn't report it. Meta's overview: About Conversions API.

There are three ways to set it up. Meta compares them in Compare Conversions API setup options:

OptionEffortBest for
Partner integration (Shopify, WooCommerce and others)A few clicksMost small stores. Start here if your platform offers it
Conversions API GatewayLow code, needs a cloud serverSites without a partner integration
Direct integrationA developerCustom sites and backends. See Get started with the Conversions API

Meta's developer guide says to use the same Pixel ID for browser and server events, and to send server events as soon as they happen, ideally within an hour (Using the API).

Deduplication is the step that matters most. If both the browser and your server report the same purchase, Meta needs to know they are one purchase, not two. Meta's recommended method: send the same event name and the same event ID from both. The Pixel passes it as eventID; the server sends it as event_id. Meta only deduplicates events that arrive within 48 hours of the first one with that ID (Handling duplicate Pixel and Conversions API events). If you skip this, your purchase count can double, your cost per purchase looks half as big as it really is, and you scale the wrong thing.

Step 3: prove it in Test Events (hours 16-20)

Don't trust a green tick. Make a real test purchase or lead and watch it arrive.

  • Open Test Events in Events Manager. Meta's guide: Test your server events using the Test Events tool.
  • For server events, send the test_event_code that the tool gives you. Remove it before you go live. Meta notes that test events are not thrown away; they still count in Events Manager (Using the API).
  • Place one test order. You should see the Purchase from the browser and from the server, with the same event ID, the right value and the right currency.
  • Check the Event Deduplication tab in Event Details. Meta says a warning appears when the deduplication rate is too low (Verifying your setup).
  • Look at Event Match Quality, a 1-10 score for how well your events can be matched to people. Meta's developer guide says it typically recommends aiming for 6.0 or higher where possible. New setups may need a little traffic before the score appears.
  • Refund or cancel the test order in your own store system so it doesn't pollute your sales numbers.

For a deeper check later, use our Pixel and Conversions API deduplication audit.

Hours 8-24: research that feeds the ads

You can't write 15 good ads without knowing what to say. Spend about a day here, in parallel with tracking.

The offer sheet (2 hours)

Write one page that answers:

QuestionExample (candle shop)
What exactly are we selling?A three-candle gift set
What does it cost, and is there a reason to buy now?$60, free shipping over $50, gift wrap included
Why should anyone believe us?1,200 reviews averaging 4.8 stars (use your real numbers only)
What's the risk for the buyer, and how do we remove it?30-day returns
What's the one sentence a customer would say to a friend?"They smell expensive and arrive gift-wrapped"

Use only claims you can prove. If a claim needs a source, write the source next to it now; the ad review and your customers will both check.

The audience notes (2 hours)

Read your own reviews, support emails and comments. Copy the exact words people use for the problem, the moment they bought and the doubts they had. Those phrases become hooks. You don't need detailed targeting for a first campaign; you need the right words.

Competitors in the Meta Ad Library (3 hours)

The Meta Ad Library shows ads that are currently running across Meta's apps. Anyone can search it, no ad account needed. Meta explains what it contains in About the Meta Ad Library.

Log 10-20 ads from three to five competitors in the workbook:

FieldWhat to write
AdvertiserName
Started runningThe date the library shows
FormatVideo, image, carousel, creator-style
HookThe first line or first two seconds
OfferPrice, discount, guarantee
AngleProblem, proof, offer or demo
Idea for usWhat you'd do differently

One warning: an ad that has run for months tells you the advertiser kept paying for it. It doesn't prove it was profitable. Use the library for patterns and gaps, not as proof of what works. Never copy another brand's creative.

If you want to pull library data with code, Meta's research team keeps a public script collection, listed in the tools section below.

Hours 24-48: the first 10-20 ads

A 4 by 4 grid: rows are four angles (problem, proof, offer, demo) and columns are four formats (9:16 video for Reels and Stories, 4:5 image for Feed, carousel, and 9:16 creator-style). Seven of the sixteen cells are marked "make 2 cuts" and nine are marked "make 1", giving 23 drafts to trim to the strongest 16 to 20.

Example mix. Vary both the idea and the shape so week one tells you which one mattered.

Why 10-20 ads, and why so different from each other? Because in week one you are not testing whether a blue button beats a red one. You are finding out which idea and which format get attention from your buyers. Ten near-identical ads teach you nothing. Sixteen ads across four angles and four formats teach you a lot.

The four angles:

  1. Problem. Name the pain in the first two seconds. "Your candles smell like nothing after 20 minutes?"
  2. Proof. Reviews, numbers, before and after. Only real ones.
  3. Offer. The price, the bundle, the guarantee, stated plainly.
  4. Demo. Show the product working: unboxing, lighting, the room filling.

The four formats and their sizes, from Meta's Ads Guide:

FormatShapeWhere it runsNote
Short video9:16 verticalReels and StoriesKeep text out of the top 14%, bottom 35% and 6% each side (Reels spec)
Static image4:5, 1440 × 1800 pxFeedMeta's recommended Feed ratio (Feed image spec)
Carousel1:1 or 4:5 cardsFeedGood for several products or a step-by-step
Creator-style video9:16 verticalReels and StoriesOne person talking to camera; a phone is fine

For every ad, fill in a row in the workbook's creative tracker: the angle, the format, the hook, the headline, the primary text, the link and a name that follows one pattern, such as angle_format_hook_v1. Our ad creative brief template helps if someone else is making them.

Add UTM tags to every link. UTM tags are labels in the link (like utm_source=meta) that let your analytics tool see which ad sent the visitor. Check them with the bulk UTM validator and keep them in the UTM tracking template.

Hours 48-60: one simple campaign

Bar chart, example arithmetic: with $50 a day and a $35 target cost per purchase, one consolidated ad set would see about 10 purchases in week one, while five ad sets at $10 a day each would see about 2 each. Same $350 and same 10 purchases in total.

Illustrative arithmetic, not a forecast. Splitting a small budget splits the signal.

With a small budget, keep the structure boring:

LevelSettingExample
CampaignObjective that matches your goal: Sales for purchases, Leads for leads. Meta's guide: How to choose an ad objectiveSales
CampaignBudget at campaign level$50 a day
Ad setConversion location and eventWebsite, Purchase
Ad setAudienceYour country, broad. Let Meta find buyers
Ad setPlacementsAutomatic (Advantage+ placements)
AdsAll 16-20 ads in the one ad set16 ads

Why one ad set? Meta's delivery system learns from results. After launch, or after a significant edit, an ad set goes through a learning phase where delivery is less stable. Meta explains this in About the learning phase. Its Marketing API reports each ad set's learning status (still learning, exited, or not getting enough results to exit) together with the conversions counted since the last significant edit, and it includes fields for a dynamic threshold (learning stage reference). So you'll often see "about 50 optimization events in a week" quoted. Treat it as a rule of thumb, not a fixed rule. Either way, $350 a week split five ways gives each ad set about 2 purchases to learn from. One ad set gets all 10. Our guide to account structure fragmentation goes deeper.

What "Advantage+" means here. Meta's developer documentation says a sales, app or leads campaign counts as an Advantage+ campaign when three settings are all on: campaign-level budget (Advantage+ campaign budget), Advantage+ audience and Advantage+ placements (Advantage+ campaign experience). For a first campaign with a small budget, that combination is a sensible default. You can still exclude existing customers or add limits later.

Before you leave this block:

  • Daily budget × 30 is at or under your spend limit, and the account spending limit is set.
  • Every ad's link opens the right page on a phone, with the price and offer the ad promises.
  • The UTM tags survive any redirects.
  • The optimization event is the one from your goal sentence.
  • The test_event_code is removed from your server events.
  • The campaign is set to start at a time someone can watch the first hours.

Our paid campaign launch checklist is a good second pair of eyes for this QA pass.

Hours 60-72: review and launch

Meta reviews ads before they run, and sometimes again after. Meta explains the process in About ads in review. You can't speed it up, so publish the campaign by hour 60 and use the wait to finish the day-1 check sheet.

If an ad is rejected, read the reason, fix the claim or image, and resubmit that one ad. Don't rebuild the campaign. Common causes are claims you can't back up, before-and-after images in sensitive categories, and landing pages that don't match the ad.

When the ads are approved and delivering, do a first readback within a few hours: is money being spent, are impressions coming in, and are new PageView events arriving in Events Manager as the clicks come in? If yes, you're live and measuring. You made it.

After launch: the day 1, day 3 and day 7 checks

Three cards. Day 1, is it working at all: ads approved and delivering, spend pacing near daily budget, events arriving in Events Manager, link clicks reach the page, spend limit set correctly; don't edit for performance. Day 3, is anything broken: which ads Meta is spending on, cost per landing page view, add-to-carts or form starts, deduplication and match quality, rejected or limited ads fixed; don't kill ads with tiny spend. Day 7, what did we learn: cost per result versus target, learning status, angles that won spend, pause ads past your loss limit, brief 5 to 10 new ads; change one thing at a time.

Each check answers a different question. Most early edits do more harm than good.

Day 1: is it working at all?

CheckWhereHealthyIf not
Ads approved and deliveringAds Manager, Delivery columnActiveFix rejected ads one at a time
Spend paceAds ManagerClose to $50 for the dayUnder-delivery on day 1 is common; wait
Events arrivingEvents ManagerPurchases or leads show from both browser and serverRecheck deduplication and the event code
Clicks reach the pageCompare link clicks with landing page viewsMost clicks become page viewsSlow or broken page. See the clicks vs landing page views gap
Safety netBilling settingsSpending limit equals your spend limitSet it now

Don't change budgets, audiences or ads on day 1 because of performance. One day of data is noise.

Day 3: is anything broken?

  • Where is the money going? Meta will spend unevenly across your ads. Some will get almost nothing. That's normal and is itself information. Our guide to new creative that gets low spend explains why.
  • Middle-of-funnel signals. Cost per landing page view, add-to-carts or form starts. These come before purchases and tell you if people are interested.
  • Measurement health. Deduplication rate and Event Match Quality in Events Manager.
  • Anything rejected or limited gets fixed now.

Don't pause ads that have spent less than your target cost per purchase. They haven't had a fair chance yet.

Day 7: what did we learn?

By day 7 the example has spent about $350. At a $35 target that is about 10 purchases if things are on target. That is enough to read the campaign, not enough to crown a single ad.

QuestionHow to answer it
Is cost per purchase near the target?Total spend ÷ purchases, from the same date range
Is the ad set still learning?The Delivery column shows the learning status
Which angles and formats did Meta spend on, and which converted?Group your ads by the angle and format in their names
Which ads hit your loss limit?Example rule: pause any ad that has spent twice the target ($70) with no purchase
What's next?Brief 5-10 new ads that build on the angles that won spend

Change one thing at a time and write it down. If you change the budget, the ads and the audience on the same day, you won't know which change did what. Our post-edit performance review shows how to read a change afterwards. If you're spending with no purchases at all by day 7, work through the spending-no-purchases diagnosis before anything else.

Budget increases are a day-7 decision, not a day-2 one. Use the budget pacing calculator to keep spend in line with the limit you set at hour 0.

The full hour-by-hour checklist

Due hourTaskDone whenMeta source
0Write the three inputs: URL, spend limit, goal sentenceGoal names one event and a target costn/a
1Create the business portfolio, turn on two-factorYou can open Business settingsBusiness portfolio
2Connect Page and InstagramBoth listed under the portfolion/a
2Start domain verificationTXT record or HTML file addedDomain verification
3Create the ad account, set time zone and currencyAccount visible in the portfolioAd account
4Add a second adminTwo people with full controln/a
5Add payment methodPayment method shows as activePayment method
6Set account spending limitLimit equals your spend limitSpending limit
8Install Pixel base codePageView arriving in Events ManagerPixel setup
12Add standard events with value and currencyPurchase or Lead fires on the real actionPixel reference
12Write the offer sheetOne page, proof for every claimn/a
16Connect the Conversions APIServer events arrivingSetup options
16Audience notes from reviews10+ customer phrases copiedn/a
18Match event IDs between browser and serverSame event name and ID from bothDeduplication
20Test purchase in Test EventsOne deduplicated Purchase with valueTest Events
24Log 10-20 competitor adsRows filled in the research tabAd Library
30Write hooks and copy for 4 angles16-23 draftsn/a
44Produce ads in 9:16 and 4:5Files meet specs and safe zonesAds Guide
48Name ads and tag links with UTMsEvery link validatedn/a
52Build campaign, ad set, adsObjective and event match the goalObjectives
58QA every link on a phoneNo broken or mismatched pagesn/a
60Remove test event code, publishAds in reviewAds in review
70Ads approved and deliveringSpend and impressions above zeron/a
72First readbackEvents from ad traffic visibleVerifying setup

The workbook turns this table into a working sheet with owners, statuses and a readiness score that only reaches 100% when the critical items are done.

When something blocks you

BlockerWhat it usually meansWhat to do
Payment declinedCard bank blocked an international chargeCall the bank, or add a second method
Domain still unverifiedDNS hasn't spread, or the TXT record has a typoRecheck the record, wait, then verify again. Don't hold the launch for it
Pixel shows no eventsCode on the wrong page, or blocked by a consent bannerUse the Pixel Helper browser extension and Test Events
Purchases counted twiceBrowser and server events without matching IDsFix eventID and event_id; see the deduplication audit
Ad rejectedA claim, image or landing page breaks a policyFix and resubmit that ad only
Account restrictedMeta flagged the account or the businessFollow the steps in Account Quality. Don't create new accounts to get around it
No spend after 12 hoursAds still in review, payment issue, or a very narrow audienceCheck review status and billing first

Tools and repos that help

Every repository below was checked on GitHub on 16 November 2026. The facebook/ business SDK links now redirect to Meta's meta/ organization.

ToolWhat it's forWho should use it
meta/facebook-python-business-sdkMeta's official Python SDK for the Marketing API, including Conversions API callsDevelopers sending server events or pulling reports in Python
meta/facebook-nodejs-business-sdkThe same, for Node.jsJavaScript and TypeScript backends
meta/capi-param-builderMeta's library that helps you collect and format Conversions API parameters correctly, in PHP, Java, Node.js, Python, Ruby and browser JavaScriptAnyone with a direct Conversions API integration who wants better match quality
facebookincubator/ConversionsAPI-Tag-for-GoogleTagManagerFiles for sending Conversions API events from server-side Google Tag ManagerTeams that already run server-side GTM
facebookresearch/Ad-Library-API-Script-RepositoryPython scripts for pulling data through the Ad Library APIResearchers and marketers who want competitor ads in a spreadsheet
pipeboard-co/meta-ads-mcpA third-party MCP (Model Context Protocol) server that lets AI assistants such as Claude read and manage Meta ads. Not made by MetaMarketers who want an AI assistant to report on the account. Start with read-only access and keep your spending limit in place

Or let GaaS do it

GaaS is built around the same three inputs: your website URL, a spend limit and a one-line goal. Its agent researches the business, makes the first ads, proposes the campaign and asks before anything spends. GaaS has made 60,000+ ads for 200+ founders. You still need the accounts, payment and tracking from the first 20 hours to exist; GaaS works inside your own ad account, under your limit.

Whether you do it yourself or not, the workbook keeps the 72 hours honest:

What this checklist doesn't cover

  • Special ad categories. Credit, employment, housing, and social issues, elections or politics have their own rules. See Meta's guide linked above.
  • Google, TikTok and other platforms. The order of work is similar, the menus and policies are not.
  • Statistical proof. A week of results on a small budget tells you where to look next. It doesn't prove one ad beats another. For real tests, use the creative testing budget calculator and the guide to budgeting a creative test.
  • Guardrails for automation. If an AI tool or agency will make changes for you, set limits first. Our guide to budget guardrails for an AI media buyer lists the ones that matter.
Set budget guardrails for an AI media buyer

Define spending authority for an AI media buyer with account scope, remaining-budget calculations, cumulative-change limits, and a reviewable approval example.

Audit Pixel and Conversions API deduplication

Trace one purchase through browser Pixel and server Conversions API events, checking shared identity, multiple senders, retries, and platform diagnostics.

Audit fragmentation in a Meta ad account

Audit whether Meta campaign and ad-set separations reflect real business constraints or unnecessarily divide budget, evidence, and operating attention.

Why a new Meta creative gets little spend

Diagnose low delivery to a new Meta ad by separating eligibility, allocation, campaign context, and the evidence required for a creative experiment.

Have a correction or a question about the workflow? Contact GaaS. Read our editorial standards for sourcing and example conventions.