Two of the numbers below came out of the same ad account in the same year. One is around $2. The other is around $120. Nothing about that business changed between them except which line of the funnel I was counting.
Founders ask me what their customer acquisition cost should be. CAC is the money it takes to bring in one new customer, and the honest response to the question is that nobody can tell you until you say what a customer is.
So here is the ladder from my own accounts. Nine prices across eight industries, every one of them a real figure from a real media buy, arranged cheapest to most expensive. The spread runs from two dollars to over two thousand.
The question that has no answer in the abstract
Search for a CAC benchmark and you land on a table. First Page Sage publishes one of the better ones, a B2C ladder compiled from 103 of their own clients, with ecommerce at $64 organic against $68 paid, and aviation at $475 against $708.
Read it closely and the thing being purchased never appears. An ecommerce customer is presumably somebody who placed a first order. An aviation customer is a seat, a booking, a loyalty account, a charter contract? The table stays quiet, because averaging across businesses that count different events cannot produce a shared unit.
The ad platforms say the quiet part out loud. Google's own documentation puts it plainly: you choose what you identify as valuable, a purchase, a sign-up, a phone call. Whatever you name, the auction goes and buys it, then reports the price back as your cost per conversion.
Every public CAC benchmark blends five definitions of customer and three definitions of cost. A ladder only means something when each rung names what was bought.
The ladder
Start at the bottom. YouMail sells spam-call blocking and voicemail handling to consumers through a mobile app. Through all of 2024 it bought an install for about $2 blended across every channel it ran.
One rung up sits a California health benefits fund. By October its bilingual campaigns were registering eligible members at about $2.40 each, down from about $35 five months earlier. Those two prices sit within forty cents of one another and share nothing else.
Registrations are where the vocabulary starts to slip. A YouMail registration, somebody creating an account, cost about $5 in the same year its install cost about $2.
Coffee Meets Bagel, a dating app whose Facebook acquisition I ran in early 2020, registered a user at about $17 in a representative week. Three registrations, three businesses, a spread of more than seven times, and one word covering all of it.
Then come the rungs where somebody actually paid. Minibar Delivery, alcohol brought from local stores to your door, took on more than 80,000 new buyers across 2020 at about $16 blended. Every Man Jack, a men's grooming brand selling direct to consumers, carried a modeled new-customer cost of about $23 for the same year.
Those two are genuinely comparable. Both mean a stranger who handed over money for the first time, in a similar price band, in the same calendar year. That is roughly the narrowest useful comparison anyone gets.
Above them the ladder stops describing consumers at all. A YouMail paying subscription cost about $120. DoorDash's New Verticals team, recruiting liquor stores and florists onto the platform, paid about $300 for a qualified merchant lead, meaning a store owner who raised their hand for a sales conversation. The ones that got all the way to a signature cost about $2,100 each.
CAC is a price, and prices only compare inside one market. An install, a lead, and a signed merchant are three different products.
Same account, different rung
The strongest evidence for that sits inside two of these accounts rather than between any of them. The $2 install and the $120 subscription are one year of one app. The distance between them is pure funnel: an install, then an account, then a free user who eventually crossed into paying.
DoorDash's pair does the same thing in a business-to-business shape at seven times the altitude. $300 gets a store owner to raise their hand. $2,100 gets one onto the platform with a signature. The 7x between them is a sales close rate expressed in money.
The same app bought an install for two dollars and a subscriber for about a hundred and twenty. Nearly sixty times apart, same account, same year. Funnel stage is the benchmark.
Which is how two companies can publish acquisition costs an order of magnitude apart while both describe a healthy business. The choice of rung gets made internally, and it almost never travels alongside the number once the number leaves the building.
It travels badly inside a company too. Change the event a campaign optimizes toward and the reported cost jumps sharply within a week, with nothing underneath it having moved. Reporting that jump as a performance change is one of the more common ways a good account gets defunded.
The other half of a definition is the cost
Every rung above is media spend divided by outcomes. None of them includes my fee, creative production, agency retainers where there were any, or the salaries of the people running the accounts.
That is the ordinary convention in paid media reporting, and it is also why a marketing CAC and a finance CAC almost never agree. On the health benefits campaign the plan carried a separate line for agency cost sitting on top of the media budget. Fold that class of cost into the denominator and a $17 blended registration becomes a different number entirely.
Check the denominator as carefully as the numerator before you compare your figure to anyone's, mine included. A fully loaded CAC can be double the same account's media-only figure while both are calculated correctly.
Two businesses that barely fit on the ladder
Rocket Mortgage puts millions of visitors a year through a home-lending funnel, and nobody there discusses acquisition in these terms. A lender's scoreboard is closed loans and loan volume, because a single closed loan outweighs a stadium full of app installs and the economics live inside the loan itself.
FX Replay, a backtesting platform for traders, has the opposite trouble. The same subscription sells into wildly different economies, and once its campaigns optimized toward paid subscriptions rather than signups, the ratio of lifetime value to acquisition cost across paid channels landed anywhere between roughly 4:1 and 15:1.
One company-wide number would have averaged that range into a figure describing none of it.
Consumer apps, public health benefits, dating, alcohol delivery, direct-to-consumer grooming, business-to-business marketplace supply, mortgage lending and trading software. Eight industries, and the last two barely put a price on a customer at all. A reference ladder earns its keep only when your business genuinely stands on one of its rungs.
How to benchmark yourself instead
The comparison that works is against your own history, one stage at a time. Twelve months of data, priced at every step: cost per click, per install where there is one, per lead or registration, per first purchase, per paying customer. Then watch each line move.
Splitting them is what makes the exercise diagnostic. A flat cost per registration beside a climbing cost per paying customer tells you the ads are working and something after the click is leaking, which is the first thing I look for in a growth audit. Both climbing together points somewhere else entirely, at the auction or at tired creative.
A single blended figure cannot separate those two situations, and they call for opposite work. One is a product and onboarding project. The other is a media and creative project.
The second comparison worth making is against your own economics rather than somebody else's average. A price is only high or low relative to what the thing is worth, which is why I run accounts on lifetime gross profit against acquisition cost instead of an industry table. $2,100 for a merchant is cheap if that merchant sells for years. $23 for a first order is expensive if the order never repeats.
What the ladder is actually for
I keep this ladder because it ends arguments fast. When somebody tells me their CAC is $40 and asks whether that is healthy, the only useful next question is which rung they are standing on.
Half the time the number turns out to be a registration, and the business has never priced a paying customer at all. The team googled an industry average, found their figure comfortably underneath it, and stopped looking.
Meanwhile the rung attached to revenue has been drifting upward for two quarters with nobody watching it. That is the specific failure the benchmark tables encourage, and it is expensive: budgets get doubled on a cheap install price and cut on an expensive lead price, when neither number was ever the one that mattered.
Name the unit. Price every stage. Compare against your own last twelve months. The benchmark you actually needed was sitting in your account the whole time.
If your acquisition numbers cannot survive the question of what they bought, this is the kind of teardown I run. Let's talk.