Blended CAC vs Channel CAC

Every store that buys traffic ends up with two acquisition cost figures that disagree, and a recurring argument about which one is real.

They are both real. They measure different things, and the disagreement between them is itself informative.

The two calculations

Blended CAC takes every dollar spent acquiring customers and divides by every new customer gained. Both inputs come from your own records: billing statements and order data. No attribution model is involved.

Channel CAC attempts the same figure per platform, using each platform’s own account of which customers it produced.

The first is verifiable. The second depends on a model that the party being measured also controls.

Why the channel figures over-claim

Platforms attribute conversions inside their own click and view windows. Meta counts a purchase it can see. Google counts the same purchase if it can see it too. Neither deducts what the other claimed, and neither knows whether the customer would have bought regardless.

The result is arithmetic that cannot be right. In the example above, the platforms claimed 830 new customers between them in a month when the store gained 640 — 190 customers who exist in reporting and not in the order table.

Because acquisition cost is spend divided by customers, an inflated denominator produces an understated CAC. Roughly thirty percent understated, in that case.

This is not deception. Each platform is answering “what did I see”, accurately, for its own window. The error appears only when the answers are summed as though they were exclusive.

What blended CAC gets wrong

Blended is not the perfect number either, and pretending otherwise leads to the opposite mistake.

It includes customers advertising did not buy — organic search, direct traffic, word of mouth, referrals. All of them sit in the denominator, so blended CAC flatters paid spend by whatever share of new customers arrived free.

That creates a specific trap during a period when organic is growing: blended CAC improves while paid efficiency is flat or deteriorating. The metric moves for reasons unrelated to the spending it appears to judge.

The fix is cheap. Track new customers from organic and direct as a share, alongside blended CAC. If the share is stable, movement in blended CAC is a paid signal. If the share is rising, the improvement may be entirely organic.

Which to use for which decision

Blended CAC for the question “can we sustain this level of spending”. It reconciles to real money and cannot be gamed, which is exactly what a decision about total budget requires. Compare it against return-adjusted contribution margin per order, and the answer is arithmetic.

Channel CAC for questions inside a channel: which audience, which creative, which campaign type. The attribution bias is consistent within an account, so relative comparisons hold even though the absolute numbers do not.

Neither for comparing one platform against another. That is the case where each platform’s independent generosity does the most damage, and where the temptation to act is strongest.

Cohorts beat months

A monthly CAC figure divides this month’s spend by this month’s new customers, and the two populations overlap imperfectly.

Spend late in a month acquires customers early in the next. In steady state that cancels out. During a scale-up it does not — spend rises before the customers arrive, so CAC looks worse while accelerating and better while slowing. Stores routinely pull back at precisely the wrong moment because of this artefact alone.

Grouping customers by first-order month and attributing the spend that acquired them removes it. Cohort CAC settles once rather than being restated every time the following month’s spending changes, and comparing the January cohort against the June one is the earliest reliable signal that a channel is saturating.

Getting the denominator right

Most attention goes to the spend side of CAC, and the customer count is where the quieter errors live.

Duplicate customer records. Guest checkout, a second email address, a different phone number. Every duplicate counts as a new customer and lowers apparent CAC. In stores that allow guest checkout this can be several percent of the count.

Repeat buyers counted as new. A customer returning after eighteen months may look new depending on how the count is built. Define “new” as no prior order ever, not no order within a window.

Wholesale, marketplace and staff orders sitting in the same table as retail. None of them were acquired by the advertising being measured.

Subscriptions. A subscription renewal is not a new customer, and treating renewals as acquisitions can transform a struggling channel into an apparently excellent one.

The check is simple: total new customers for a year should reconcile against the growth in your unique customer count over the same period. If it does not, the definition is leaking somewhere.

The gap as a diagnostic

The difference between blended and summed channel figures is worth watching in its own right.

A stable gap means your attribution is consistently generous by a known amount, which is manageable. A widening gap usually means either more channels are claiming the same conversions, or organic and returning-customer revenue is growing — and those have opposite implications for how hard to push paid spend.

Neither figure alone shows that. The relationship between them does, which is the practical argument for calculating both rather than picking a side in the usual argument.

Track the ratio of summed channel claims to actual new customers as a standing number. In the example it is 1.30, and what matters is not the level but whether it holds steady. A stable ratio means your channel figures are wrong by a predictable factor you can mentally adjust for. A moving one means the adjustment you have been applying no longer holds.