Cost per Customer Acquisition: What the Number Includes and Where It Misleads

A board pack puts customer acquisition cost at $940. The finance model for the same quarter puts it near $2,600. Nobody made an arithmetic mistake. The two calculations counted different costs in the numerator and different customers in the denominator, and neither one wrote down the choices it had made. The board version was paid media divided by every new customer. The finance version was fully loaded sales and marketing divided by the customers that sales and marketing could plausibly claim.

Both numbers are defensible. Only one of them answers whatever question is about to be decided, and picking the wrong one is how companies conclude that a channel is cheap right before they scale it into unprofitability.

The formula, and the two arguments hiding inside it

Cost per customer acquisition is the total cost of acquiring customers over a period, divided by the number of new customers acquired in that period. The arithmetic is trivial. Every real disagreement about CAC is an argument about one of two inputs: what belongs in the numerator, and which customers belong in the denominator.

Neither of those has a single correct answer. What matters is that the answer is written down, applied consistently, and reported alongside the number rather than reconstructed from memory six months later when someone asks why CAC moved.

What belongs in the numerator

The instinct is to count media spend. Media spend is usually the smallest part of the real cost.

Everything below is an acquisition cost in the sense that it would fall away if you stopped acquiring customers, and most companies exclude at least two of them.

Fully loaded people cost. Salary, employer taxes, benefits and equipment for everyone whose job is winning new customers: SDRs, account executives, demand generation, content, marketing operations, the paid media manager. Loaded cost runs meaningfully above base salary, and using base salary alone understates the biggest line in the numerator.

Commission and variable compensation. Paid on closed business, so it belongs to the customers it closed. Companies that exclude it are usually excluding it because it lands in a different part of the P&L, which is an accounting reason rather than an economic one.

Agency and contractor fees. Retainers, per-meeting or per-lead fees, freelance writers, design, video. This line is often the easiest to find and the most commonly reported in isolation, which is part of why agency-sourced CAC looks flattering when nothing else is loaded next to it.

Tooling. The sales engagement platform, the CRM seats used by acquisition roles, enrichment and data providers, sending infrastructure, call recording, scheduling, the intent data subscription somebody renewed. Individually small, collectively a real line.

Content and creative production. Cost of producing the asset, not just distributing it. A piece of content that keeps producing pipeline for two years is a genuine argument for amortising it, and that argument needs to be made explicitly rather than by quietly leaving the cost out.

The debatable ones are worth naming too. Customer success and onboarding usually sit in cost of goods sold rather than acquisition, though the boundary blurs when onboarding is doing the closing. Brand spend with no attributable pipeline is still acquisition cost in the sense that it buys future customers, and companies that exclude it are choosing to report a cleaner number rather than a complete one. Executive time is real and almost never counted.

CAC numerator: include or exclude

• ✓Yes: Fully loaded salaries for sales and marketing headcount
• ✓Yes: Commission and variable compensation on new business
• ✓Yes: Agency retainers, per-lead and per-meeting fees
• ✓Yes: Paid media and sponsorship spend
• ✓Yes: Acquisition tooling: CRM seats, sending infrastructure, data providers
• ~Depends: Content and creative production cost
• ~Depends: Brand spend with no attributable pipeline
• ~Depends: Executive and founder selling time
• ✕No: Onboarding and customer success delivery
• ✕No: Expansion and upsell cost against existing accounts

The denominator is less obvious than new logos

New customers acquired in the period sounds unambiguous until you try to count them.

Expansion revenue from existing accounts is not acquisition, so upsell should leave both sides of the ratio. Customers who signed and churned inside the period still cost money to acquire, so removing them flatters the number. Customers who came through a partner or a referral programme with its own cost line need that cost in the numerator if they are in the denominator.

The sharpest version of the problem is self-serve alongside sales-led. A company with a free tier that converts people who never spoke to a human, and an enterprise motion with a nine-month cycle, has two businesses with different unit economics. Dividing the combined cost by the combined customer count produces a number that describes neither. Segment first, then divide.

Blended, paid and channel CAC

Three versions of the number get used interchangeably, and they answer three different questions.

Blended CACAll cost, all customers

• Every acquisition cost divided by every new customer
• Answers: what does growth cost the business overall
• Right for board reporting and cash planning
• Flattered by organic, referral and word of mouth
• Useless for deciding where the next dollar goes

Paid CACPaid cost, paid customers

• Paid acquisition cost divided by customers from paid
• Answers: what does buying a customer cost right now
• Right for budget and scaling decisions
• Depends entirely on attribution being honest
• Rises as you scale, because the cheap audience goes first

Channel CACOne channel at a time

• Cost and customers for outbound, paid search, events, partners
• Answers: which channels deserve more budget
• Right for mix decisions and quarterly reallocation
• Hardest to compute, because cost allocation is judgement
• Breaks down when channels assist each other

Blended CAC is the number most often quoted and the least useful for making a decision. It includes customers who arrived because a founder is well known, because an existing customer recommended you, or because you have ranked for a category term for four years. Those customers cost something, and it is nothing like the cost of the next one.

Why blended CAC flatters strong inbound

Consider the mechanics rather than a specific company. A business with a large organic and referral base has a denominator inflated by customers that no amount of extra spend would have produced faster. Divide total spend by that larger denominator and the average falls. The average falls even if the paid motion is losing money on every customer it buys.

The failure follows predictably. Blended CAC looks comfortably below the target, so the paid budget is doubled on the strength of it. The organic base does not double. The average moves toward the marginal cost of the channel that actually scaled, and the number that looked healthy in January is unrecognisable by June. Nothing degraded. The mix changed, and blended CAC was always a mix statistic wearing the clothes of a unit cost.

The practical discipline is to report blended and paid side by side, permanently. The gap between them is a measure of how much of your growth you are not paying for directly, which is useful in its own right, and it makes the flattery visible instead of load-bearing.

The lag problem

CAC divides a cost incurred in one period by customers who closed in the same period. In a business with a short cycle that is roughly fair. In a business with a six-month cycle it is a category error, because this quarter's customers were bought by last quarter's and the quarter before's spend.

The effect is systematic rather than random. Grow spend and reported CAC overstates true cost, because the numerator has scaled while the denominator still reflects the smaller earlier spend. Cut spend and reported CAC understates it, which is how a company convinces itself that a budget cut improved efficiency. The most common form of this is a hiring ramp: new reps carry full cost from month one and close nothing for a quarter, so CAC deteriorates on paper during exactly the period the investment is being made.

Two fixes, both cheap. Use a trailing window matched to your sales cycle length rather than a calendar quarter. And run a cohort view alongside it, holding spend against the customers that spend actually produced, even when they close two quarters later.

1. Step 1Segment first

   Split self-serve from sales-led, and new business from expansion. A combined number describes neither motion.

2. Step 2Load the numerator

   Fully loaded people cost, commission, agency fees, media, tooling. Decide the debatable lines explicitly.

3. Step 3Match the window

   Use a trailing period matched to your sales cycle rather than a calendar quarter, so lag does not fake a trend.

4. Step 4Report three versions

   Blended, paid and per channel, side by side. The gaps between them carry the information.

5. Step 5Write down the choices

   Record what was included and excluded next to the number, so the next calculation is comparable rather than reinvented.

Payback period is the companion metric

CAC on its own says nothing about whether a cost is affordable. A high CAC against a high margin annual contract can be excellent. A modest CAC against a thin margin product that churns inside a year can be fatal.

Payback period closes the gap: CAC divided by monthly gross profit per customer, expressed in months. It answers how long the business is out of pocket before a customer has repaid what it cost to win them. Gross profit rather than revenue is the part people get wrong, because revenue payback ignores the cost of serving the customer and produces a number that is optimistic by exactly the amount of your cost of goods sold.

Payback is the more actionable of the two because it is a cash statement. It tells you how much working capital growth consumes, which is the constraint that actually binds most companies, and it responds to margin improvement as well as to cost reduction. The ratio of lifetime value to CAC is worth watching too, with the caveat that lifetime value contains a retention assumption and therefore contains an opinion. Payback contains almost none.

For the same reason, payback tolerance should come from your own cash position and funding situation rather than from a published benchmark for your category. A company with two years of runway and a company with seven months can look at identical unit economics and correctly reach opposite conclusions.

Where the number misleads even when it is calculated correctly

It is an average over a mix. A change in CAC frequently means the mix moved rather than that any channel got better or worse. Always ask whether the components moved before congratulating anyone on the average.

It hides the marginal cost. The relevant number for a budget decision is what the next customer costs, not what the last hundred averaged. Marginal cost rises with volume in every paid channel, because the most responsive audience is reached first. Average CAC can fall while marginal CAC is already above what you can afford.

One-off costs distort short windows. A conference sponsorship, a website rebuild or a severance payment lands in one month and produces customers over many. Amortise or annotate, and never read a single month as a trend.

Attribution decides channel CAC before you do. Under last touch, the channel that closed gets everything. Under first touch, the channel that found the prospect does. The same quarter can produce very different channel CAC depending on a setting in an analytics tool, which is worth remembering before reallocating budget on the strength of it. The relationship between measured cost per lead and eventual cost per customer is where this bites hardest, and our breakdown of what a lead generation agency costs covers how the upstream number distorts on the way down.

Definition drift. A number that includes contractor spend one quarter and excludes it the next produces a trend that exists only in the spreadsheet. This is the most common cause of a CAC improvement nobody can explain.

Making it defensible

Write the definition down once and put it next to the number every time it is reported. Segment before dividing. Report blended and paid together so the mix is visible. Match the window to the sales cycle. Pair CAC with gross margin payback so affordability is part of the same conversation. Recalculate the same way every period, and treat any change to the definition as an event worth annotating.

The number is a management tool rather than a score. Its job is to make the next allocation decision better than the last one, and it can only do that when everyone reading it is looking at the same construction.

When you are comparing what an internal team, an agency retainer and an outsourced motion actually cost per customer, the loaded comparison rarely matches the quoted one. Our breakdowns of lead generation agency cost and outsourced SDR versus in-house work through the lines that are usually missing on one side, and the MQL and SQL distinction matters here because a CAC built on a soft stage definition inherits that softness all the way down.

If you would rather see the cost per customer of a live outbound motion in your market before modelling it, you can see what a campaign would look like.