CPA (cost per acquisition)

CPA is advertising spend divided by conversions: the cost of one purchase, lead, or signup. On its own the number carries no information about whether it is good. A $40 CPA is cheap for a business with $65 of contribution margin per order and ruinous for one with $28, so CPA is only ever meaningful against an allowable figure derived from your own economics.

For ecommerce that allowable figure is contribution margin per order, and the arithmetic is direct: a $150 order at 43.8% margin gives $65.65 of contribution, so $65.65 is breakeven CPA and the target sits below it by whatever profit you intend to keep. For subscription and lead-gen businesses the ceiling comes from payback instead — how much you can pay for a customer and still recover it within the number of months your cash position tolerates.

The larger problem with CPA is not the target, it is the definition of the conversion being counted. Whatever event you name, the platform will optimize toward it, so the choice of event decides what the algorithm buys. Name a soft event and you will get volumes of it.

This is why CPA on raw leads is among the most gameable numbers in marketing. Work it through: 500 leads at $30 gives a $15,000 spend and a CPA that looks excellent. If 4% of those leads close at $2,000, that is 20 customers and a real acquisition cost of $750. A different channel delivering 200 leads at $60 — twice the reported CPA — closing at 14% produces 28 customers at $429. Judged on lead CPA the first channel wins comfortably. Judged on customers it is nearly twice as expensive, and shifting budget toward it makes the business worse while the dashboard improves.

The response is to optimize to the deepest event that still has enough volume for the algorithm to learn from, and to value leads by what they become rather than counting them. Feeding qualified-lead or closed-revenue signals back to the platforms changes the population it pursues; counting all leads equally guarantees it pursues the cheapest ones, which are cheap for a reason.

Two further notes. Read CPA alongside volume, because a falling CPA achieved by retreating into your warmest audience is a shrinking business with an improving metric. And keep the conversion definition stable, since redefining the event mid-quarter makes the trend line meaningless and resets what the bidding has learned.