CPM, CPC and CPA: Reading the Numbers

A cost metric expresses spend relative to a count of outcomes. The formula can be straightforward, but the meaning depends on what the spend includes and how the outcome is counted.

CPM: cost per thousand impressions

CPM equals spend divided by impressions, multiplied by 1,000. With $500 media spend and 100,000 impressions, CPM is $5.

The number describes the price of delivery under the chosen definition. It does not by itself show whether the ads were viewable, suitable for the brand or effective for the advertiser.

CPC: cost per click

CPC equals spend divided by clicks. At $500 spend and 250 clicks, CPC is $2.

Check which clicks the platform reports. An ad click, a destination visit and an analytics session are different events. Redirects, filtering and measurement conditions can prevent these counts from aligning.

CPA: cost per acquisition or action

CPA equals spend divided by qualifying conversions. With $500 spend and 10 qualifying purchases, CPA is $50 per purchase.

Define the action before comparing campaigns. A registration and a purchase should not share a generic conversion label without explanation. Attribution windows and deduplication rules also affect the count.

Use a consistent cost basis

One report may contain media cost while another includes platform fees. Align cost basis, currency, dates and outcome definitions before comparing them.

When there are zero outcomes, the ratio is undefined. Report spend and zero outcomes rather than inventing a zero cost per outcome.

Aggregate the inputs

To combine campaigns, add spend and outcomes, then calculate the ratio. An unweighted average of CPAs can mislead when one campaign is much larger.

For example, $100 divided by one action and $900 divided by 30 actions gives a combined CPA of about $32.26: $1,000 divided by 31. Averaging $100 and $30 incorrectly gives $65.

Answer the actual question

Choose a metric that fits the objective and pair it with quality and business context. A lower CPC is not automatically better if the traffic is less useful. Report what happened after the interaction and the measurement limitations as well as the cost ratio.