Cost per acquisition, not certified public accountant
In advertising, CPA stands for cost per acquisition: the amount of media spend it takes to buy one conversion. Divide what a campaign spent by the number of conversions it recorded and you have it. Most searches for those three letters are after a certified public accountant instead, which is an accountancy qualification and entirely unrelated. The operative detail for a buyer is that acquisition means whichever event you told the platform to optimise toward, so two figures only compare when they count the same action.
The nearest confusable neighbours live one letter away. Cost per click prices traffic. Cost per lead prices a contact detail. Cost per acquisition prices the event you actually wanted, which is usually a sale. And customer acquisition cost is a business figure that carries salaries, retainers and production, so it always lands higher than anything the ad account shows.
The number a media buyer optimises toward
This is the figure a buyer watches hourly. It sets the target in campaign setup, it decides which ad sets get more budget by lunchtime, and it is the trigger for pausing an ad. What makes that harder than it sounds is noise. At fifty dollars a purchase on a hundred dollar daily budget, one extra sale swings the reported number by a third. Small spend produces confident-looking figures that mean very little.
Then there is the window. Spend two thousand dollars, record forty purchases, and you report fifty dollars a purchase. Switch the attribution setting from seven day click to one day and some of those forty vanish, because the same campaign now gets credited for less. Nothing about the advertising changed. Treat the window as part of the metric's name and write it into the report, or your month-on-month comparison is measuring a settings change.
Why platform and finance figures disagree
Ask the ad account and ask the finance lead and you get two different answers, both honest. The platform counts conversions it believes it caused, inside its own window, using its own pixel data. Finance counts orders that cleared in the bank and divides by everything spent to get them: media, agency retainer, creator fees, editing, samples and the discount codes. The second answer is always the larger one.
The gap widens when several channels claim the same sale. Run Meta and TikTok together, add email, and the conversions credited across those dashboards can add up to more orders than the store actually processed. So keep a blended check running: total spend over total new orders for the month, taken from your own store data. Use the platform view to choose between ads and the blended view to judge whether the programme works.
Setting the number you will allow
Work backwards from margin instead of borrowing a target from a blog post. Take your average order value, subtract cost of goods, shipping, payment fees and an allowance for returns, and what remains is the most an order can cost before the sale stops paying. Set your allowable figure below that, and leave headroom for the weeks when impression prices climb. If you prefer thinking in multiples, the same margin gives you a break-even return on ad spend.
Then use the allowable figure as a gate on creative, not just on budget. Rank every creator video by cost per purchase over a fair window, keep the ones sitting under the line, and commission variations of whatever those did. Give each one enough spend to produce a believable count of sales first. Pulling a video after two days because three purchases arrived slowly is how a decent creator gets dropped for no reason.
The other half of the gate is what the asset cost to make. A video that only just meets your allowable figure after four revision rounds and a steep licence renewal is worth less than one that hit the same mark on a flat fee. Keep production cost per usable asset in the same sheet as cost per purchase, and the ranking of your creators quietly rearranges itself.
How it's used
Cost per purchase on the bundle creative is holding near forty dollars on a seven day click window, but the blended view says closer to sixty, so I am not scaling it yet.
Just so the scope is clear, we pay on delivery of the video and not on the sales it drives, so there is no performance component in this agreement.