Context
Artopad sells 3D puzzles. When the work started the product had never been sold anywhere: no audience data, no pixel history, no benchmark for what a customer should cost. The owner had roughly $10,000 to cover the fee, the media, and everything else. The brief wasn’t "grow." It was closer to "find out whether this sells, and don’t spend the whole budget finding out."
Diagnosis
At that budget Meta was the only channel that could produce enough conversions to learn from in the first few weeks. Everything else would have produced opinions.
The website was the weaker part of the system and that was flagged before launch. The client wanted to test the product on the existing site first, which was a fair position, so month one became a test of the product rather than the funnel.
The data settled one question early. Most people bought more than one unit, and the two-pack was the common order. Order value, not the price on the box, was the number to hold against acquisition cost.
Decisions
- Released the budget in stages against evidence instead of committing it up front
- Ran a separate clicks campaign to fill the audience pool, excluding anyone already reached, and closed those people in remarketing
- Put the US and Canada in one campaign with separate ad sets: similar CPM, similar behaviour, and splitting them would have fragmented the signal
- Kept the UK separate, because the market behaved differently
Results
Roughly 800 orders and 2,000 units across three months, with returns under 1%. The inventory sold out. Revenue and gross contribution are intentionally not published because the available figures are calculated from price architecture rather than reconciled against the store back end.
The strongest month ran 377 purchases on $2,798.61 of spend, a blended $7.42. Video performed best, particularly footage of people handling the product. Conversion improved once the client shipped the site changes.
The engagement ended because the stock sold out, and by then the China tariffs had made restocking uneconomic.