CPA Marketing: How Cost Per Acquisition Works in Paid Media

CPA (Cost Per Acquisition) in paid media is the amount spent to generate one desired action: an install, a registration, a subscription purchase, or any other event you define as a conversion. If a campaign spent $5,000 and drove 250 app installs, the CPA is $20 per install.

The "acquisition" in CPA can refer to many different actions depending on what you're optimizing for. Install CPA, purchase CPA, and subscription CPA are all valid uses of the term. When someone says CPA without specifying the event, they usually mean the primary conversion event in the campaign.

CPA Bidding: How It Works

When you set up campaigns using CPA-based bidding, you tell the ad platform a target CPA — the amount you're willing to pay per conversion — and the platform adjusts bids in real time to try to hit that target. This is called target CPA (tCPA) on Google and cost cap bidding on Meta.

The algorithm works by predicting which users are most likely to complete the desired action at the lowest cost, then bidding accordingly. A user with a high probability of converting gets a higher bid (because acquiring them is efficient). A user with low probability gets a lower bid or no bid at all.

For this to work well, the platform needs sufficient signal. Google's tCPA typically requires at least 30-50 conversions per month in the account for the algorithm to optimize meaningfully. Meta performs better with higher volumes. Below these thresholds, you're better off with manual bidding or maximum delivery (let the platform spend freely and measure CPA manually).

The Limitation of CPA Bidding

CPA bidding optimizes for the conversion event you defined — it doesn't optimize for the value of that conversion.

If you're bidding on installs, the algorithm finds users who will install. But a user who installs and churns in 24 hours costs the same CPA as a user who installs and subscribes for two years. The algorithm doesn't know which you got until well after the fact.

This is why BeFreed switched from tCPA (optimizing for installs) to value-based bidding (passing subscription revenue signals back to the platform) on Google UAC. After the switch, CPA for installs increased slightly — but ROAS improved 38% because the algorithm was now finding users who actually paid, not just users who clicked.

The same principle applies to any funnel. If you optimize for installs but 80% of your LTV comes from the 10% of users who subscribe, targeting installs broadly is inefficient. Targeting the behavior pattern associated with subscribers (even if you're tracking installs as the conversion event) gets you closer to the users who matter.

CPA vs. CAC

These terms are used interchangeably by some teams but measure different things.

CPA is a campaign metric. It measures what you paid on a specific ad platform for a specific conversion event, using the platform's own attribution. A $20 CPA on Google means Google is crediting your campaign with an install at a $20 cost based on their attribution model.

CAC is a business metric. It's the total cost to acquire one paying customer when you include all sales and marketing spend, divided by total new customers. CAC is almost always higher than CPA because it includes overhead, and because some campaigns generate installs that never convert to paying customers.

If your Google campaign shows a $20 install CPA but only 15% of installs become paying subscribers, your effective CAC from that campaign is closer to $133 per subscriber — even though Google is reporting $20 per install.

When to Use CPA vs. ROAS as Your Optimization Goal

Use CPA bidding when:

• Your product has low price variation per user (everyone pays the same amount)
• You're in early campaigns without enough purchase data for value bidding
• You're optimizing for a non-revenue action (registrations, trials)

Use ROAS or value-based bidding when:

• Users vary significantly in purchase value
• You have sufficient conversion data (50+ purchases per month at minimum)
• You want to optimize for revenue rather than conversion count

AIMA monitors CPA performance across campaigns and automatically tests bidding strategy shifts (from tCPA to value-based and back) when volume thresholds are met, so you're not running the wrong bidding strategy as your campaign matures.

Managing the Learning Phase

CPA bidding works best when the platform has enough conversion signal to optimize delivery efficiently. During the learning phase, the initial period after launching or significantly changing a campaign, performance is often inconsistent and CPA targets are frequently missed.

For Meta, the learning phase exits when an ad set achieves roughly 50 optimization events per week. For Google tCPA, around 30 to 50 conversions per month at the account level. Below these thresholds, the algorithm is effectively estimating, and CPAs may run 30 to 50% above your target before settling.

Common mistakes during the learning phase: reducing the CPA cap because initial performance looks poor, making frequent bid or budget changes (each significant change resets the learning clock), and running too many ad sets with small individual budgets so none reaches the conversion threshold.

A better approach: launch with a CPA target set 20 to 30% above your real goal to give the algorithm room to learn. Let it run for at least 7 to 10 days before making adjustments. Once the campaign exits learning, tighten the CPA cap gradually toward your actual target rather than jumping directly to it.

CPA by Funnel Stage

CPA applies across the full funnel, but the right conversion event to optimize against depends on where most of your actual LTV comes from.

Install CPA is the easiest to hit because it has the highest volume and lowest intent threshold. It is also the most disconnected from business outcomes. An install CPA of $3 looks efficient until 80% of installs never open the app a second time.

Trial or registration CPA is a better proxy for monetizable users in apps with free trials or freemium models. Users who complete onboarding and reach a trial state signal genuine product intent. Optimize against this event when your install-to-registration rate is sufficient (typically 20% or higher) to give the algorithm enough signal.

Purchase or subscription CPA is the most accurate optimization event because it directly targets revenue-generating users. It requires enough monthly conversions for the algorithm to optimize effectively, so it typically only works once the campaign reaches at least 50 monthly purchases.

The right progression: teams optimizing for installs should move to trial-start optimization when volume allows; teams optimizing for trial starts should move to subscriptions when volume allows. Each step up the funnel gets the algorithm closer to the users who generate real revenue, even though CPA figures increase at each step.