How to calculate new-customer CAC (and why your ad platform gets it wrong)

How do I calculate what it really costs me to win a new customer, and why does my ad platform show a lower number?

Customer acquisition cost is one of the few marketing numbers a finance team trusts on sight, which is why getting it wrong is expensive. The usual error is not the formula, which is a single division. It is what goes into it: which spend, and above all which customers count as new. This guide gives the formula, a worked example, a template to copy, and the three reasons the cost per new customer your ad platform shows is not your CAC.

The formula

New-customer CAC = acquisition spend in the period ÷ new customers acquired in the period.

Both halves need a definition, written down once and applied every month.

• Acquisition spend is what you paid to win customers: ads aimed at people who have not bought yet, including retargeting of visitors who never purchased. Ads aimed at existing customers are retention and belong in a separate line.
• Paid or fully loaded. Paid CAC counts media only. Fully loaded CAC adds agency fees, creative production and tools. Both are legitimate; mixing them in one trend line is not.
• New customers are people whose first settled order in your store's entire history falls inside the period. Not their first order this year, and not their first order in the report's window: their first order ever.

A worked example

An illustration with round numbers. In one month a store spends $150,000 on ads, $90,000 on Meta and $60,000 on Google. 1,200 customers buy that month: 700 for the first time and 500 who had bought before.

Dividing by every customer gives $125 and looks healthy, but it spreads the cost of acquisition over 500 people who were not acquired this month. The real figure is $214: the $125 hides 42% of what each new customer cost. And the platforms, added together, report 900 new customers for a store that gained 700, which makes acquisition look cheaper again.

Why your ad platform calculates it wrong

A platform's cost per new customer is a reasonable number for comparing campaigns inside that platform. As the store's CAC it is wrong, for three reasons that all push it the same way: down.

It uses its own definition of new

Google's customer acquisition goal, by default, builds its list of existing customers from up to the last 540 days of campaign activity and tracked purchases, and Google's own help page notes that this method is less comprehensive because people remove cookies and opt out. Meta's new-customer settings rely on the audiences you tell it are your existing customers. Either way, a buyer from three years ago, or one whose email the platform never matched, is new to the platform and returning to your store. We explain the three competing definitions in why platforms disagree about who is a new customer.

It counts the new customers it credits to itself

Each platform divides its own spend by the conversions it attributes to its own ads, under its own windows, view-through included. A first-time buyer who clicked a Meta ad on Monday and a Google ad on Thursday can be counted by both, which is how 700 real new customers become 900 reported ones.

It does not know what happened to the order

A first order that was cancelled or refunded in full still counts for the platform. For your store it produced no customer, and counting it lowers CAC with a sale that never settled.

A template you can copy

One row per month, filled from your store's orders, never from the ad platforms. The example month from above is filled in.

• New customers: customers whose earliest settled order is in the month, matching the same person across guest checkouts and accounts by email and customer ID at least.
• First-order gross profit: first-order revenue minus the cost of the products, and ideally minus shipping and payment fees. At a 35% margin, a $1,050 first order leaves about $368.
• First order pays back CAC: yes when the first order's gross profit is at least the CAC. When it is not, the customer has to buy again before the acquisition breaks even, which is the question lifetime value answers.

Rules that keep the number honest

• Compute first-order status over the whole order history. A bounded window turns every seasonal customer back into a new one each time their last purchase ages out of it.
• Leave out first orders that were cancelled or refunded in full, and apply the same rule in every report. We cover why no platform subtracts them in refunds and cancellations: the gap no ad platform subtracts.
• Keep one definition of spend for the whole series, paid or fully loaded, and write it on the report.
• For CAC by channel, divide each channel's spend by the new customers whose first order your own attribution credits to that channel, not by the new customers the platform reports.
• Read it over a quarter before acting on it. A single month moves with promotions, seasonality and one large campaign.

What to compare it with

CAC on its own cannot say whether a customer was worth acquiring. Compare it with the gross profit of the first order, which says whether the first purchase pays for the acquisition, and with the gross profit the customer produces over time, which we cover in LTV in ecommerce. For the channel view that sits next to it, see ROAS, POAS and MER. Adray marks every order as a new or returning customer from the store's full order history, matching the same person across customer ID, email and phone, which is exactly the denominator this formula needs.