THE CURRICULUM / CH 3 — Marketing Mastery

3.1Facebook Setup

A build order you run once, in sequence, and never revisit in a panic.

5 min read

It is the Thursday before you launch and you cannot verify your domain, because the domain sits under an old email address, which is attached to a personal account you last logged into years ago, which is the only admin on the page you need. None of this is advertising. All of it stands between you and spending a dollar.

Set once, live with forever

The decisionWhy it follows you
CurrencyEvery report from that account is denominated in it. Pick the one you bank in.
Time zoneEvery day boundary and budget reset inherits it. Pick the one you live in.
Whose portfolio it sits inAssets made inside somebody else portfolio can be taken back.

the first two cannot be changed afterwards. the third can, but not by you.

TWO OF THESE CANNOT BE CHANGED AFTERWARDS.

the third can, but not by you

THE BUILD ORDER

The three layers of an ad account drawn as nested blocks, with the decisions each one owns. The campaign owns the objective, the bid or buying strategy, the budget if set at that level, and the attribution setting for the report. The ad set owns who it can be shown to, where it is allowed to appear, the optimisation event it learns on, the schedule, and the budget if set there instead. The ad owns the creative itself, the destination it sends people to, and the post the comments accumulate on. No budget, bid or result figure is shown.
The three layers you are building, and the decisions each one owns.
  1. Business portfolio under your own login. Two-factor authentication on.
  2. A second admin. One owner, one phone, one point of failure.
  3. Create or claim the Page. A stranger sees it at the decision.
  4. Ad account inside the portfolio. Currency, time zone, payment method, backup.
  5. Install the pixel. Confirm in the test tool that events fire in order.
  6. Add the server-side stream, sending the same events from your backend.
  7. Place one real test order. Confirm the purchase appears once, not twice.
  8. Verify the domain, or you do not control how your events are ranked.
  9. Rank priority events, purchase at the top. Anything above it hides purchases.
  10. Business verification now. Mid-appeal is two hard things at once.
The five funnel stages from impressions to purchases, each with a narrowing bar, and between them what a drop at that step points at: the hook or audience, a page that does not match the ad, the offer or price or proof, and checkout or shipping or trust. No conversion rates are shown.
The order the events fire in, and what a drop at each step points at.

They fail in different places

Browser pixel

  • Sees on-site behaviour your backend never handles.
  • Stripped by blockers, limited by tracking prevention, lost when the tab closes.
  • It never reports an error. It reports a smaller number.

Server-side stream

  • The same events from your server. Meta calls it the Conversions API.
  • Survives blockers, and carries the permitted match parameters that tie an event to a person.
  • Knows only what your backend knows.

DOUBLED NUMBERS ARE THE MOST EXPENSIVE KIND OF GOOD NEWS.

no shared event id, and every decision rests on a number you did not earn

One purchase reported twice and counted once. The browser pixel is JavaScript in the buyer browser and sees on-site actions your backend never sees, but it loses blocked scripts, tracking prevention, a tab closed early and a session that moved. The server stream sends the same events from your own backend and survives blockers, but it loses anything the backend never learns about. Both reports carry the same event name and the same event ID, which is what deduplicates them, so the platform counts one. Without the shared ID it counts two. No count or rate is shown.
The shared event ID is the only thing stopping one purchase counting twice.
One purchase, counted by two different ledgers. Two things decide whether the platform records it. First the window: how long after the ad a purchase still counts, so a purchase made inside the window is counted and one made after it closed is not. Second the device: whether the half that saw the ad and the half that bought can be joined up, so a purchase on the same device is joined up and one on another device may or may not be. The platform counts events; your bank counts money. No figures appear.
The platform counts events. Your bank counts money.

A policy rejection is a data point, not a coin flip. Read the reason, appeal rather than repost, and change the thing it named. The ad and the landing page are reviewed as one, so a claim you cannot substantiate on the page is not a claim you have.

The pages a store must have before it can take money, with two columns of marks showing which of the two inspections checks each one: the payment processor and the ad reviewer. Product and checkout, contact, shipping policy, returns and refunds, terms of service and privacy policy are marked for both. About and FAQ are marked for neither, and exist for the buyer. Each row also states what its absence signals. The marks are a checklist, not a score.
The reviewer reads the page as well as the ad. These are the pages it looks for.

Renting an ad account

What it buysWhat is left out
A support contact who can escalate, not an appeal form.Access, not ownership. Withdrawable, with no appeal.
Sometimes higher spend limits from the start.You inherit the neighbourhood. Another renter can take you down.
A way to keep running while your own account is restricted.A fee on every unit of spend, and whatever restricted you stays undiagnosed.

rent only against a problem you tried to solve directly: an appeal you lost, or limits you reached.

ACCESS, NOT OWNERSHIP.

withdrawable, with no appeal

A nine-box tick list an ad must clear before it is allowed to spend, drawn as a form with a hairline box beside every line. The pixel fires in order and is confirmed in the test tool rather than assumed. The server-side stream sends the same events from your own backend. One real test order shows the purchase appearing once, not twice. The domain is verified and priority events are ranked with purchase at the top. Every claim in the ad is substantiated on the page it lands on. The question is written in one sentence and the variable that moves is named. Success and kill criteria are written down with metric, window and date. The spend comes from your own break-even cost per purchase rather than from ambition. Variants differ in exactly one thing and are named to explain themselves. The verdict rule at the foot is absolute: if any box is unticked, the ad does not spend. Under it, the obligation after launch — nothing is edited inside the window, because every edit restarts the learning clock.
What has to be true before an ad is allowed to spend.
Three tiers of number ranked by how much you can trust them. Measured: money that landed in the bank, orders in your admin, invoices you paid. Modelled: platform-attributed conversions, anything with a lookback window, blended estimates. Unknown: what you assumed, what you were told, what you hope. This is a rank, not a score.
Measured, modelled, unknown. A rank, not a score.

TAKEAWAYS

  • Build inside a portfolio you created and administer. Anything inside another one is withdrawable.
  • Currency and time zone are set once. Every report that account produces inherits both.
  • Verify the domain and rank priority events before you spend, not after.
  • Policy rejections are diagnostic. Read the reason, appeal, and change what it named.

CHECK YOURSELF

Why is a server-side event stream not simply a replacement for the browser pixel?

The clearest structural risk of renting an agency ad account is:

DO THIS BEFORE THE NEXT MODULE

  1. 01Run the build order end to end, ticking each step off as you finish it.
  2. 02Add a second admin and a backup payment method to the account today.
  3. 03Complete business verification now, while nothing of yours is under appeal.
Enrol free