Cost per Acquisition (CPA) is the cost an advertiser pays each time a user completes a specific conversion action - typically a purchase, but sometimes a signup, trial, or download. In the CPA model, you pay a set fee per conversion rather than per click or per impression. This makes CPA a performance-based metric: you only pay when the desired outcome happens.
CPA is related to but distinct from Customer Acquisition Cost (CAC). CAC is the all-in cost of acquiring a new customer across all channels and time periods - it includes ad spend, agency fees, content costs, and discounts. CPA is typically channel-specific and campaign-specific: the cost of a conversion from a particular Meta campaign, Google Shopping campaign, or affiliate partner. CAC aggregates all of those CPAs (and non-paid acquisition costs) into a single business-level figure.
CPA and ROAS measure campaign efficiency from opposite directions. CPA asks: how much did it cost to generate one conversion? ROAS asks: how much revenue did each dollar of spend generate? Both are valid, but they surface different insights. A campaign with a low CPA but low average order value may have poor ROAS even though the individual conversion cost looks healthy. A campaign with high ROAS but high CPA may be profitable on a per-order basis but unsustainable at scale if the customers acquired have low lifetime value.
A target CPA should be derived from your unit economics, not set arbitrarily. The upper bound for a sustainable CPA is determined by your gross margin, average order value, and target LTV:CAC ratio. A brand with 60% gross margin and a $90 AOV has roughly $54 in gross profit per order - meaning a CPA above $54 destroys margin on the first transaction. Factoring in a 3:1 LTV:CAC target and a 24-month customer lifespan produces a much higher allowable CPA, but requires confidence in the lifetime value projections underpinning that calculation. Using A/B testing and channel-level CPA tracking in tools like Triple Whale or Northbeam allows Shopify brands to continuously optimise toward their target CPA without relying on platform-reported attribution alone.
.png)
.png)
