Why does our customer acquisition cost look fine until you separate new customers from repeat ones?
Blended CAC divides all marketing spend by all jobs, including repeat customers who would have called anyway. That flatters the number. New-customer CAC divides the same spend by first-time customers only, and it is almost always the higher figure. Both are legitimate. Blended tells you what marketing costs per job today; new-customer CAC tells you what growth actually costs.
Two denominators, two very different numbers
Blended CAC is total marketing spend divided by total jobs in the period. New-customer CAC is the same spend divided by jobs from customers who had no prior record with you. The numerator is identical. Only the denominator changes, and the denominator is where the argument lives.
The gap between the two is a direct read on how much of your work is being carried by the base you already built. A large gap is not a problem. Not knowing the size of the gap is.
Why the gap widens as a company matures
A five-year-old company with a large maintenance base will show a blended CAC that improves every year even if its acquisition machinery has gotten worse. Repeat jobs pile into the denominator, the average drops, and nobody notices that first-time customer counts have been flat for eight quarters.
The early warning is the ratio of first-time customers to total customers served. Track it monthly. When it drifts down while blended CAC also drifts down, that is not efficiency. That is the base absorbing the shortfall. Customer intelligence exists partly to make that drift visible before it shows up in revenue.
Which number to use for which decision
Use blended CAC when you are asking what marketing costs the business as a whole, or comparing marketing spend to total revenue on a profit and loss statement.
Use new-customer CAC for every budget decision. If you are deciding whether to add spend to a channel, the only relevant question is what it costs to produce a customer who was not already yours. Channel comparison built on blended numbers systematically favors whichever channel repeat customers happen to route through, usually branded search or direct.
The identification problem underneath both
Neither number is better than your definition of new. Most field service systems will happily create a duplicate customer record when a person calls from a different phone number, or when the address is entered with a different suffix. Every duplicate inflates your new-customer count and makes acquisition look cheaper than it is.
Before trusting either metric, run a deduplication pass on phone, address and email, and decide a lookback rule: is someone new if their last job was three years ago? Write the rule down. Identity resolution is the unglamorous half of marketing intelligence, and it is where most of these numbers break.
Topics: CAC · new customers · repeat customers · metrics
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.