Reading
Customer acquisition cost
Acquisition spend divided by the customers it brought. The arithmetic takes a second; everything that makes the number worth anything is in what you put on top of the division, and in knowing which of the two things that moved it actually moved.
The formula
CAC = acquisition spend ÷ new customers, measured over the same period for both. That is the whole calculation. Spend in one month often buys a customer in the next, so businesses with a long sales cycle lag the numerator deliberately — but the shape does not change.
If you want it done on your own figures, including the part below about spend that reached nobody, the CAC calculator does the division and shows its working.
The numerator is the argument
Nobody disputes the division. What people disagree about is the first term, and the gap between the two common conventions is larger than most of the performance differences anyone is trying to measure.
Paid media only
Ad spend and nothing else. Easy to pull, easy to compare between channels, and it flatters you — it leaves out everyone whose salary made the campaign exist. This is the useful number for deciding where the next dollar goes.
Fully loaded
Media plus salaries, tools, agency fees and creative. Closer to what the customer actually cost, much harder to assemble, and the number a board means when it asks. Always larger, by however much of your payroll points at acquisition — which is a figure only you have.
Worked: $40,000 of media and 160 new customers in a month is 40,000 ÷ 160 = $250. Add $20,000 of salaries and tools and the same month is $375. Same customers, same campaigns, one number 50% larger than the other. That is the entire reason to say which one you are quoting.
The denominator has one trap
Count customers acquired in the period, not customers who happened to pay in it. A renewal is not an acquisition, and counting it as one makes a bad month look like a good one — reliably, and in the direction nobody checks.
What a good CAC is
There is no benchmark worth quoting in isolation, because the number only means something against what a customer is worth. The two that get quoted most come from the same place — David Skok’s SaaS metrics — and both are repeated more strictly than he wrote them.
On the ratio he says the best SaaS businesses run “higher than 3, sometimes as high as 7 or 8”. On payback, that the best “recover their CAC in 5-7 months”, and that past twelve months profitability becomes anemic. Which makes 3:1 a floor rather than a target, and twelve months the point where it starts to hurt rather than the goal — two distinctions that disappear every time the figures are restated, including in the first version of this page.
They are also one practitioner’s numbers for subscription software. If you sell something else they are a starting point and not a verdict, and the comparison that always works is against your own last quarter: a CAC that moved twenty per cent in a month with nothing changing in the market is worth explaining whatever the ratio says.
Why it moved, and the answer nobody checks
CAC is a fraction, so it rises for two reasons that look identical in a report: the top got bigger, or the bottom got smaller. Everyone investigates the top. Auction prices, creative fatigue, seasonality, a competitor bidding harder — all real, all well covered, all addressed by people whose job is the ad account.
The bottom is where the cheap explanation lives. Conversions stop when the page the ads point at stops working, and the version of that failure which costs the most is the one that still answers 200 OK: a page telling visitors it does not exist while every monitor watching the status code reports it healthy. Spend continues. Customers do not arrive. CAC rises, and reads in the report exactly like an auction problem.
That failure is unusual in being total rather than partial. Most marketing inefficiency is a matter of degree — this audience converts a little worse than that one — and moves the fraction a little. Spend that reaches a destination which cannot convert anybody brings no customers at all, so it adds to the numerator and contributes exactly nothing to the denominator. It moves CAC by its whole value.
In INC-2291 that was $12,480 at $58.59 a minute across 3 campaigns before anybody acted — and every dollar of it landed on CAC as pure numerator. Checking whether it is happening to a page right now is free and takes about three seconds.
What CAC does not tell you
Worth saying, because a metric quoted without its limits is how teams end up optimising the wrong thing:
- It says nothing about whether the customer was worth acquiring. A falling CAC that is buying worse customers is a worse business, and the number improves.
- It hides the mix. One blended figure across channels can fall while every individual channel gets more expensive, if a cheap channel grew.
- It lags. Spend attributed to the month it was spent, customers to the month they closed — on a long sales cycle those are different cohorts wearing the same date.
- It cannot tell you which half of the fraction moved, which is the thing you actually need to know, and is why the destination is worth watching separately from the spend.
The failure modes behind the second half of this page are published as a CC BY 4.0 dataset — every way a paid destination stops working, which HTTP status each returns, and what an uptime monitor reports while it happens. Five of the six still answer 200 OK. Quote it with attribution; it includes what we do not detect.