Ask three people in the same business what their customer acquisition cost is and you will get three defensible answers. A media buyer divides platform spend by platform-attributed purchases. The finance director adds agency fees, creative production and software, then divides by first-time buyers only. Somebody in the middle blends every channel together and produces a third figure. None of them is lying. They are counting different things, and nobody writes down which version they used.
This page is for operators who need one figure they can defend in a board meeting and still bid to on a Tuesday morning.
What customer acquisition cost actually includes
Everything you spent to win a customer, divided by the number of customers you won. Every argument about this metric is really an argument about one of those two halves.
The spend half should carry media, agency or freelancer fees, creative production, and any software that keeps the campaigns running. Add first-order discounts, because a welcome code funded out of margin is acquisition spend wearing a different coat.
The customer half should carry first-time buyers. Not orders, not conversions, not sessions.
Most reporting gives you neither. It gives you platform spend over platform-attributed purchases, which is a useful operational figure and a weak business one.
So I hand over both. One version tells a media buyer whether to raise a bid tomorrow. The other tells the owner whether the business is making money, and the two answers disagree more often than anybody expects.
The costs that never make it into the numerator
Four of them, and it is the same four almost every time.
Agency or freelancer fees come first, because they are usually the largest single addition. Creative production is second, covering video, static variants, and the people who make them. Software is third: feed tools, analytics, and whatever you pay to move conversion data between systems. Discounts are fourth, and they hide best of all.
A welcome code never appears in any ad account. It still bought the customer, and it came out of the same margin the campaign is being judged against.
Shipping subsidies behave identically. Free delivery on a first order is acquisition spend that lands in the fulfilment budget.
None of this changes what you bid tomorrow. It changes what you believe about the business. I have worked on accounts where the platform figure looked comfortable against margin and the full figure did not, and nobody had ever put the two side by side.
Orders are not customers, so the denominator is yours
Divide by orders and you have measured cost per order. Divide by first-time buyers and you have measured acquisition. Those are different numbers in any store where people come back.
Both of the dated figures I publish are order-based, and I label them that way every time, because the ad platforms could not tell me how many distinct people sat behind those orders.
The gap widens exactly when retention is working. A store with strong repeat buying shows a flattering cost per order and a much harsher customer acquisition cost, since the repeat orders were nearly free to generate.
Your commerce platform knows who has bought before. Ask it for the new customer count and use that as your denominator, rather than accepting whatever the ad account calls a purchase. A blended acquisition cost built from total marketing spend over that same store-side count is the version a finance team will recognise.
Two acquisition costs I can date and count
€14.18 across 1,715 orders, and $5.05 across more than 63,000 purchases. Both cover the full year 2022, and that date travels with them wherever they go.
The first came from an EU fashion ecommerce client running Meta and Google together, with the majority of sales arriving through Google Search and Shopping. Presenting it as a Meta result would be wrong. The second came from a US electric mobility ecommerce client, on Google Search, Performance Max and Display.
Roughly three times the difference in nominal terms, and both businesses were healthy at the time. Different products, different price points, different margins.
Neither figure is a target for 2026. Auction prices have moved a long way since, and Meta’s own investor reporting put average price per ad up 12% year on year in both the first and the second quarter of 2026.
I publish the volume and the date because almost nothing else written on this subject does either.
Customer acquisition cost at account level hides the spread
One live account, one period, four honest answers.
A US flag and banner ecommerce client, Meta sales campaigns, June to August 2026. Across all six sales ad sets the cost per purchase was $67.24 on 64 purchases. Excluding one failed test it reads $53.72 on 61 purchases. My best ad set bought orders at $29.37, and the worst active one paid $93.96.
Every one of those is accurate. None of them is the account’s customer acquisition cost, because all four count orders rather than people, and none carries my fee.
That is the ordinary state of ecommerce reporting. Pick a level, write down which one you picked, and stay on it. A figure that quietly drifts between account average and best campaign will show you an improvement that never happened.
Customer acquisition cost in lead generation needs a close rate
In lead generation the same phrase means something else, and a close rate stands between the two.
I held £11.40 across 6,070 leads for a UK staffing and childcare client, on Google Search only. That is a cost per lead. Turning it into an acquisition figure needs the share of those leads that became placements, which lives in the client’s system rather than in mine.
Other accounts sit nowhere near it. A US auto accessories client produced 906 leads at $91.18, on Search and Performance Max with roughly 80% of the spend on Search. A US auto transport client on Meta produced 5,599 leads at $2.68, held steady while spend scaled, having previously sat above $6.
My cheapest lead and my dearest one differ by more than thirty times. For scale rather than as a target, WordStream’s 2026 study of 13,474 US search campaigns put the all-industry average at $66.69, covering April 2025 to March 2026.
A broken conversion event invents a customer acquisition cost
One account in my portfolio reports an acquisition figure that is pure fiction, and its traffic side is completely ordinary.
It is a US automotive equipment ecommerce account. Delivery reads normally: $8.28 CPM, $0.28 CPC, a 2.92% click-through rate, and 43,415 link clicks. Nothing there raises an eyebrow.
Then the conversion side claims a return on ad spend of 44.37. No ecommerce account on earth produces that. What it produces is a purchase event firing more than once, or firing on the wrong page, or being counted twice by two systems.
Every customer acquisition cost calculated from that account is wrong by an unknown multiple. So before you compare yourself to anybody, place one test order and confirm the event fires once. Broken measurement will happily show you a spectacular number while the bank balance says otherwise.
Setting a ceiling from margin instead of a benchmark
Start with contribution, because that is the only ceiling you can defend.
Across my ecommerce clients gross margin runs between 20% and 50%, which puts break-even return on ad spend between 2.0x and 5.0x. Take a real basket size to make it concrete. Shopify reported an average order value of $114.70 across its full merchant base for Black Friday Cyber Monday 2025, published on 2 December 2025, which is first-party data from the platform itself.
At 40% margin that order contributes $45.88. So $45.88 is the absolute ceiling on the first order, and the ceiling for media alone sits lower again once fees, creative and tooling come out.
At 20% margin the same order leaves $22.94. Push margin to 50% and it leaves $57.35.
Identical revenue, ceilings differing by more than double. That is why an industry average with no margin attached cannot tell you whether you are winning.
What to change this week
Three things, in this order.
Write down which version of the number you are quoting, using two lines: what went into the spend, and what went into the count. Most disputes about performance turn out to be two people quoting two different calculations at each other.
Then get the new customer count from your commerce platform rather than the ad account, and divide total marketing spend by it. That single figure is the one your accountant will recognise.
Then set your ceiling from contribution rather than from a benchmark table, and take fees out before you hand a target to whoever buys the media.
Why the lead version and the purchase version are not the same metric covers the close rate problem in detail. What an online order actually costs me at volume gives both dated figures with their sample sizes attached. Where the three to one rule really came from explains why a borrowed ratio is worth less than your own margin, and how long the money takes to come back turns that ratio into a calendar. Worked examples from real accounts sit in my case studies. If your own figure will not survive scrutiny, book a teardown.