CAC Is Not CPA, and Confusing Them Costs Real Money

In most marketing meetings, CAC and CPA are used as synonyms, and the sloppiness is not free. I have watched a board approve an aggressive growth plan built on a $180 "CAC" that was actually a platform-reported cost per attributed conversion, and watched the same board's face eight months later when the fully loaded number turned out to be $560. Nothing about the marketing had changed. Only the honesty of the arithmetic had. So before the formula, the distinction: CPA is a campaign metric, the cost of a conversion event inside a channel, computed by the channel. CAC is a business metric, everything you spend to win a customer, computed by you. One belongs in a media dashboard. The other belongs in your unit economics, your fundraising deck, and your pricing. Confuse them and every downstream decision inherits the error.

The honest CAC formula is simple to state and uncomfortable to apply: all sales and marketing costs in a period, divided by new customers acquired in that period. The discomfort is in "all." Media spend, yes, but also agency and tool fees, the salaries of the people running the programs, sales compensation for the humans who close what marketing sources, creative production, the discounts and free-trial subsidies that purchased the first order. Most published "CAC" figures include media and nothing else, which is why founders comparing their numbers to benchmark charts are usually comparing their partial number to other people's partial numbers, a comfort ritual with no information in it. Compute it fully loaded, before anyone builds a plan on it, the sequencing argument of stop presenting media plans before you know the margin, and separately compute paid CAC, media divided by customers from paid channels, because the two numbers answer different questions: the first tells you whether the business model works, the second whether the media does.

Two subtleties fix most broken CAC math I encounter. First, the numerator and denominator must live in the same time zone: this quarter's spend produces some of this quarter's customers and some of next quarter's, so businesses with long sales cycles should lag the denominator or compute CAC by cohort, spend in the month a cohort was sourced against customers that cohort eventually produced. Second, new customers means new; blending reactivations and renewals into the denominator is the quiet way growth teams flatter themselves, and it dissolves the moment anyone checks.

Now the question the searcher actually has: is my CAC good? On its own, the number is meaningless, and every "average CAC by industry" table proves it by being useless. CAC is only ever good relative to what a customer is worth, which makes the working ratio LTV to CAC, lifetime value against acquisition cost. The folk threshold of 3-to-1 is a reasonable starting posture for subscription businesses, but the ratio is only as honest as the LTV inside it, and LTV built on optimistic retention curves is how companies buy growth that later evaporates, the pattern I documented in the subscriber you buy on discount leaves on schedule. For non-subscription businesses I prefer a blunter instrument: months to payback. Contribution margin per customer per month against CAC tells you how long your money is out the door, and whether you can afford your own growth rate at your current cash position. A company with a 3:1 LTV:CAC and a 30-month payback can be simultaneously healthy on paper and dead in practice.

The cohort version of the calculation is worth showing once, because it is where long-cycle businesses stop lying to themselves. A B2B client sourcing pipeline in Q1 spent $310,000 on sales and marketing that quarter and closed 14 new customers inside it, a naive CAC of $22,000 that had the board alarmed. Cohorted honestly, that Q1 spend eventually produced 41 customers as its pipeline matured over the following two quarters: true cohort CAC, $7,600, comfortably inside their allowable. The same error runs the other direction in companies that are slowing down, where today's closes were bought by last year's bigger budget and the naive math flatters a program already in decline. Divide same-period numbers in a long-cycle business and you will always be wrong; the only question is which direction.

The last trap is the one that ruins good companies quietly: managing CAC as an average instead of at the margin. Your blended CAC includes the customers who arrived nearly free, brand search, referrals, organic demand your product earned. Averages built on that base make every incremental paid dollar look affordable long after it stopped being so, because the marginal customer, the one your next dollar buys, always costs more than the average one. This is the same shrink-to-glory geometry I described in what is a good ROAS, running in the opposite direction: ROAS flatters you into spending too little, blended CAC flatters you into spending too much. The fix for both is identical, measure the margin, and the tooling is the unglamorous kind, spend-differential tests and the holdout designs from you do not need a data science team to run a holdout, which convert CAC from a reported average into an observed cost curve.

So: calculate it fully loaded, cohort it if your cycle is long, ratio it against an LTV you would defend under oath or a payback period your balance sheet can carry, and watch the marginal number, not the blend. CPA tells you what a channel charges for an event. CAC tells you what growth actually costs. Companies that keep the two words separate tend, not coincidentally, to keep their margins too.