Ongoing case study

$357 in Meta ads. 33 sales in week one.

A $67 online course, launched on Meta ads from zero. These are the real numbers from the first seven days, pulled straight from Meta Ads Manager and the store's checkout. I'll keep adding weeks as it runs.

Rio CostaFounder, Uprise. I run this account. Updated Oct 4, 2026

Blended = every sale in the store that week divided by ad spend. It includes a few sales from her own bio link, so the Meta number is the conservative one. Both are below.

The client

Maya BC Films (bonhamcarterfilms.com, @mayabcfilms) is a filmmaker and editor who teaches editing. Her product is the Edit Bootcamp: six recorded calls, about nine hours, her sound effects pack included, $67 one time.

Before this, she sold it through a link-in-bio store. She wanted to own the page and the buyer data, and put paid traffic behind it.

It's not a clothing brand. It doesn't matter. A $67 course and a $60 hoodie run through the same Meta machine: one product, a checkout, a purchase event, creative that gives people a reason to buy. Every rule in why clothing brand ads stop working came from running this account.

What we built before spending a dollar

1. A sales page on her own domain. One product, her reels, real student reviews, a $67 button. Checkout runs through Whop.
2. A members library where buyers log in and watch the recordings, plus an admin panel so she can add or reorder lessons herself.
3. Pixel plus Conversions API. The browser pixel fires the purchase, and the server sends the same purchase to Meta with a matching ID so it's counted once. Buyers on iPhone or inside the Instagram browser still get credited to the ad.
4. A sales board that joins Meta spend to real checkout sales per ad, so we judge ads on money in the account, not on what Meta guesses.
5. Creative from her own reels. I recut three reels she'd already posted, stripped the "comment for the link" endings, and added one end card with the offer. Plus two statics: a timeline and a real student review (used with his permission).

We tested a $0 purchase end to end before launch and confirmed Meta received it. No spend until that passed.

Week one, day by day

Source: Meta Ads Manager, account level, Sep 27 to Oct 3, 2026. 18,034 impressions.

Look at Tuesday. Cost per purchase went from $6.68 to $23.76 overnight. That's the day most people kill the campaign or start editing it. We didn't touch it, and the next three days added 10 more Meta purchases. One bad day is not a dead ad.

Week one, ad by ad

Ad D ran as two copies; their numbers are combined. Our checkout tracking also caught a sale on ad C that Meta didn't attribute.

What week one taught us

• One ad carries the account. Ad A spent 68% of the budget and drove 74% of Meta purchases. Meta found it fast and kept feeding it. That's normal, and it's why you test several and let Meta pick.
• Creative fatigue isn't close yet. Frequency is 1.5 on the winner. Refresh starts when it climbs toward 3, so new ads go in next to A, never as edits to A.
• Video beat statics on volume. The statics bought sales cheaply when they got delivery, but Meta barely spent on them. Native-looking video from her own feed did the heavy lifting.
• Server-side tracking paid for itself on day one. The first sale didn't show up in Ads Manager until the Conversions API sent it. Without it, Meta would have optimized blind.
• Budget only goes up in steps. $25 a day to start, then raises of about 20% at a time so Meta's learning phase doesn't reset.

How the numbers are counted

Two sources, both shown, nothing rounded up.

• Meta purchases (27): what Meta attributes to the ads. This is the number you'd see in Ads Manager.
• Store sales (33, $2,211.01): every paid order in checkout that week. 14 carried an ad tag straight through checkout, 4 came from her own bio link, and 15 had no tag (usually someone who saw the ad on their phone and bought later in a different browser). Some of those 15 are ad sales Meta also counted, so the two numbers overlap and you can't add them.

If you only trust one number, use the Meta one: $13.25 per purchase on a $67 product.

Watch me walk through the account

I filmed this on day six, when the account was at $307 spent and 24 purchases. Full breakdown: the 5 reasons clothing brand ads stop working.

Week two

In progress. New creative goes in next to the winner, budget keeps stepping up, and I'll post the numbers here the same way: spend, purchases, cost per purchase, straight from the account.

Questions