Alcohol rules kept the catalog out of Meta. The checkout sat on a partner's domain. We built the page and the tracking first, then scaled.

The Situation

Coppercraft is a craft distillery in Holland, Michigan, founded in 2012, with a tasting room, a nine-year straight bourbon, and a rye that has taken Double Gold. In the glass, the product does the selling. Online, almost none of that carried over.

Three things stood in the way. Meta treats alcohol as a restricted category, so the product catalog could not run as a normal shopping feed, every ad had to clear a 21+ and ship-state gate, and every claim on an image had to survive content review. Purchases completed on a licensed third-party checkout on another domain, so the pixel could not see the sale it had created. And the brief was explicit: the owner wanted purchases and a cost per purchase, not impressions or engagement.

What We Did

We did not start with ads. We started with the page and the tracking, because a working ad pointed at a page that cannot convert or report is just an expensive way to find out nothing.

• Landing pages built for cold traffic, one per bottle. Age gate, ship-state logic, the age statement and proof front and center, awards cited by name, and a buying path that hands off cleanly to the licensed checkout. When the 9-Year sold out, the page became a restock list instead of a dead end.
• Tracking through the partner checkout. Purchase events wired so Meta could see completed orders on the other domain. Every purchase figure on this page comes from that pipe.
• Angle testing, then concentration. Roughly fifteen small tests across the three bottles. One angle converted profitably: nine years in the barrel at $37.99, against the $80 most nine-year bourbons ask. In the first thirty-day review it produced 62% of all purchases. Spend moved behind it. The rest was cut.
• Compliance built into the creative pipeline. Every concept clears the alcohol checklist before render: no one who reads under 25, no consumption cues, no unsubstantiated superlatives, proof and age statements as the claim.
• Weekly review, one change at a time. Scaling in 20 to 30% steps every three days, no new ads dropped into a learning ad set, and every change justified by a number.

The Results

Across the life of the account, the purchase campaigns have produced just over a thousand Meta-attributed orders at an average cost of about $18 each, and a 3.3x return on ad spend. Cold acquisition has been the engine. In the first six weeks the store recorded 259 orders and a 14.5% conversion rate on paid traffic, while the sales campaign ran at $16.70 per purchase.

Two things did not work, and we said so. The retargeting pool, run as one undifferentiated audience, hit a frequency above five and produced a worse cost per order than cold traffic. It was restructured, and the worst ads in it were switched off. And the winning 9-Year angle did not transfer to the Blend. A direct port of the same copy converted nothing, because the Blend buyer's objection is youth, not price. Each bottle now gets its own argument.