The Customer Acquisition Cost Formula, and Why the Denominator Breaks First

Your CAC went from $40 to $48 last month. Nobody changed a bid, no auction got dearer, and the creative is the same creative. Somebody is now going to spend a week finding out what went wrong, and the answer is that nothing went wrong. You scaled spend, your customers arrive on a lag, and the formula divided one month's larger budget by a customer count that was still half-built from the smaller budget before it.

Almost every guide to the customer acquisition cost formula spends its length arguing about the numerator. Do salaries count. Does the CRM subscription count. What about the designer who works on acquisition two days a week. Those arguments are real but they are second order, because a numerator decision is at least stable once you make it. The denominator changes underneath you every month, and it changes most when you are doing the thing you want to be doing, which is spending more.

The formula, and the two arguments it starts

The formula is not in dispute:

CAC = acquisition cost over a period / new customers acquired in that period

Take a month where you spent $40,000 on paid media, $4,000 on agency retainer and tools, and $16,000 on the loaded salaries of the two people who do nothing but acquisition. That is $60,000. A thousand new customers signed up. Fully loaded CAC is $60.00. Media-only CAC is $40,000 / 1,000, which is $40.00, so the fully loaded number your finance team will use is fifty percent higher.

Both numbers are correct and they answer different questions. Media-only CAC tells you whether the next dollar into the auction is worth putting there, because salaries do not rise when you raise a budget. Fully loaded CAC tells you whether the business is viable, because salaries are real money leaving a real bank account. Pick one for bidding decisions and the other for board decisions, write down which is which, and stop converting between them mid-sentence. Most CAC arguments inside a company are two people using different numerators and neither saying so.

What belongs in the numerator

Two tests settle most of it. First, would this cost fall if you switched the ads off tomorrow? If yes, it is acquisition cost. Agency fees usually pass. Office rent does not. Second, is this cost spent on people who are not yet customers? Your retention emails, your loyalty programme and your support desk all fail that one. They are real costs and they belong in your contribution margin, not in CAC. Putting them in CAC double counts them, because they are already sitting inside the contribution margin you are about to compare CAC against.

The honest awkward case is the salary of somebody who does half acquisition and half retention. Split it by time, write the split down, and never revisit it mid-year. A CAC series is only useful if the rule that produced it stayed still. A better rule applied from March onwards makes January and March incomparable, which costs you more than the worse rule did.

The denominator moves your CAC when nothing else did

Here is the arithmetic that produces the phantom spike in the opening paragraph. Assume your true cost to acquire a customer is a flat $40.00 and never moves. Assume half the customers a month's spend buys sign up in that month, and half sign up the month after, which is unremarkable for anything with a considered purchase.

Month one, you spend $40,000. It buys 1,000 customers: 500 land this month, 500 land next. Another 500 land this month from the month before it, so you report 1,000 customers and a CAC of $40.00.

Month two, you scale to $60,000. That buys 1,500 customers: 750 land this month, 750 land next. Add the 500 still arriving from month one and you report 1,250 customers. Your reported CAC is $60,000 / 1,250, which is $48.00.

Month three, you hold spend at $60,000. Now 750 arrive from month two and 750 from month three, so you report 1,500 and a CAC of $60,000 / 1,500, back to $40.00.

Month four is the part that should worry you more than month two. You cut spend back to $40,000. It buys 1,000, of which 500 land now, and 750 are still arriving from month three, so you report 1,250 customers against $40,000 and a CAC of $32.00. Cutting a third of your budget made your acquisition look twenty percent cheaper than it has ever been. Reported CAC ranged from $32.00 to $48.00, a spread of fifty percent, while the true number sat at $40.00 the entire time.

The fix is not clever. Stop dividing this month's spend by this month's customers and start dividing a cohort's spend by the customers that cohort produced, measured once the lag has run out. If you do not have cohort data, the cheap approximation is a three month rolling CAC, which will not remove the distortion but will stop you acting on it. Either way, the rule is that you never compare a rising-spend month to a falling-spend month and call the difference performance. Post a warning in whatever dashboard your team reads, because the flattering version in month four is the one that gets quoted upward.

Blended CAC is not an average

The second denominator problem is customers you did not pay for. Say paid media brings 1,000 customers at $60,000, so paid CAC is $60.00, and another 500 arrive through search, referral and direct. Blended CAC divides $60,000 by 1,500 and returns $40.00.

Nothing about that is wrong, but it is not an average of two channel CACs and it does not behave like one. Watch what happens when paid works and you double it. Paid spend goes to $120,000 and brings 2,000 customers, still at $60.00 each, so paid efficiency has not moved by a cent. Organic is still 500, because organic does not scale because you increased a budget. Blended CAC is now $120,000 / 2,500, which is $48.00. Your blended number rose twenty percent as a direct consequence of paid holding steady and succeeding.

Blended CAC is a useful number for one question only, which is what it costs the whole business to add a customer right now. It is the wrong number for every decision about the auction, and it is a dangerous target, because the fastest way to improve it is to cut the paid spend that was funding your growth. Keep both, label both, and never set a bonus against the blended one.

The number CAC has to beat

CAC on its own means nothing. It only becomes a decision when you put it against contribution, which is what is left of an order after the cost of the goods, the payment fee, the shipping and the returns. Across this site the worked example runs on an $86.00 average order and a 42 percent contribution margin, which is $36.12 an order.

Against that, a fully loaded CAC of $60.00 loses $23.88 on the first order. Two orders give you $72.24 of contribution against $60.00 of CAC, so you clear $12.24, and the whole business case rests on the second order arriving. The flattering media-only figure does not rescue you either: $40.00 against $36.12 is still $3.88 short of a single order. This is the thing the formula will never tell you, and it is why CAC payback period and not CAC is the number worth putting on a wall. Run your own figures through the CAC calculator and then the payback period calculator, in that order.

It also explains why cost per acquisition and CAC are not the same measurement even though the arithmetic looks similar. CPA is denominated in whatever the platform counted as an action, which is usually a first order and sometimes a lead. CAC is denominated in customers, and a customer is a thing that can come back. If your CPA and your CAC are identical, you are either a business with no repeat purchase or you are counting something wrong.

Why CAC benchmarks travel badly

A CAC of $600 is comfortable for a business whose contribution per customer is $2,000, because it costs thirty percent of what a customer is worth. A CAC of $40 is fatal at our $36.12, because it costs about 111 percent of it. There is no dollar figure that is good, only a ratio, and the ratio is set by your margin and your repeat rate rather than by your industry. Benchmarks collected across companies with different margins are measuring the margins, not the acquisition.

The same caution applies to every CAC figure you read, including the ones on this site. Treat published benchmarks as a check that you have not misplaced a decimal point, not as a target. Your own break-even, built from your own contribution margin and your own repeat rate, is the only ceiling that applies to you, exactly as it is for CPM and every other cost metric.

Do one thing this week. Pull the last six months of spend and new customers, compute CAC the same-period way and then again on a three month rolling basis, and put the two lines on one chart. The months where they diverge are the months your team made a decision about a number that was not real. If you want the fuller version of the argument, the lifetime value calculator is where the repeat rate that sets your CAC ceiling gets built.

Related reading: how CAC differs from CPA and the LTV to CAC ratio.