CAC (customer acquisition cost)

The denominator problem: who counts as a new customer

CAC is a division, and almost every serious error in it lives below the line. The numerator is at least anchored to invoices. The denominator is a definition, and if nobody has written that definition down, the reporting layer has silently chosen one.

The specific ways it goes wrong are consistent across brands. Guest checkout with a different email address creates a new customer record for someone who has ordered four times. Reactivated lapsed customers are counted as acquisitions, which lowers CAC without a single new person entering the business. A second subscription from the same household counts twice. Subscription renewals are sometimes counted as orders from new customers because the renewal creates a new order row. Wholesale and marketplace orders land in the same table as direct ones and inflate the count for a channel that had nothing to do with them.

All of these move CAC in the same direction — down — which is why they survive. A definition that makes the number worse gets questioned; one that improves it rarely does. The fix is unglamorous: define "new customer" once, in the data layer, on a stable identity such as a resolved customer record rather than an order row, and make every report read that definition rather than reimplementing it.

Blended, paid and channel CAC

Take an example month: $100,000 of total marketing cost, of which $80,000 is media spend and $20,000 is agency fees, creative production and tooling. The business acquires 1,000 new customers, and the ad platforms between them claim 700 of those.

What belongs in the numerator

Media spend is never in dispute. Everything else is, and the useful test is whether the cost would fall away if you stopped acquiring customers. Agency retainers, creative production, affiliate commissions, influencer fees and the cost of gifted product all pass that test. Rent does not. Fully loaded salaries for an in-house acquisition team pass it in principle and are excluded in practice by most brands, which is defensible as long as it is consistent and stated.

The cost most often missed is the acquisition offer itself, because it sits in a different line of the P&L. Continue the example: if each of those 1,000 new customers redeemed a 20% welcome discount on an average first order of $80, that is $16 per customer and $16,000 in the month. Adding it moves blended CAC from $100 to $116 — sixteen dollars a head of real acquisition cost that never appeared in any ad account, and that grows every time the welcome offer is deepened to hit a volume target.

Free shipping on a first order behaves identically. So does the cost of a sample or trial size given away to open a relationship. None of these are marketing spend in the accounting sense and all of them are acquisition cost in the economic sense.

Consistency matters more than getting the boundary theoretically right. A CAC that changed because the definition changed is the single most common cause of a metric improving while nothing improved, so the boundary should be written down and moved deliberately, with the history restated when it moves.

LTV:CAC and the horizon caveat

The folklore target is 3:1. It is quoted without two qualifications that decide whether it means anything.

The first is that LTV has to be contribution-margin LTV, not revenue. Gross-revenue LTV ignores cost of goods, fulfilment, processing and returns, and on a thin-margin catalogue it will justify an acquisition cost that loses money on every customer acquired. A 3:1 ratio on revenue at a 30% margin is really 0.9:1 on margin, which is a business paying for the privilege of growing.

The second is that the ratio contains no time. A 4:1 ratio realised over three years and a 4:1 ratio realised in four months are the same number and completely different businesses, and only one of them can be funded from cash flow. This is why payback period belongs next to the ratio rather than underneath it.

The practical discipline is to quote a horizon every time. "12-month contribution LTV to blended CAC of 1.8" is a claim someone can check and act on. "LTV:CAC of 4" on a two-year-old business is an extrapolation of a curve nobody has observed to its end, and it will be extrapolated optimistically because the person doing it wants to spend more.

One last mismatch worth naming: comparing a blended LTV, which includes organically acquired customers who tend to retain better, against a paid CAC. The numerator and denominator then describe different populations, and the ratio flatters paid acquisition by exactly the amount organic customers are better.