Customer acquisition cost (CAC) explained — and how to lower it

Customer acquisition cost is total sales and marketing spend divided by the number of new customers acquired in the same period. That formula is the easy part and it hides the hard part: CAC is an accounting choice before it is a number, and four defensible sets of choices applied to the same quarter will produce four different CACs, none of them wrong. The useful work is deciding your boundaries, writing them down, keeping them stable, and then reading the number against payback rather than against a ratio. This is how to do each of those, with the arithmetic worked through.

CAC is an accounting choice before it is a number

Customer acquisition cost appears in no accounting standard. A public company files sales and marketing expense; CAC is a ratio somebody builds on top of that using assumptions they rarely publish. That has two consequences worth internalising before you compute anything.

Cross-company CAC comparisons are close to meaningless. A competitor quoting a CAC of $400 against your $1,100 may be excluding salaries, or counting a different denominator, or amortising content spend across four quarters. You cannot tell, and neither can they.

Your own CAC is only useful as a time series. The absolute value matters far less than the direction, and the direction is only readable if your definition did not change between periods. The single most common way teams break this metric is redefining it — usually by adding a cost category — and then reading the resulting jump as a business event.

So the first task is not calculation. It is writing a one-paragraph definition, dating it, and treating a change to it as a change that requires restating history.

The formula, and the four boundaries people actually disagree about

CAC = sales and marketing cost for a period / new customers acquired in that period

Written as a sentence, since that is how it will be read: take everything you spent on acquiring customers over a period and divide it by how many you got. Four decisions determine what "everything" means.

Boundary one — people costs. Fully loaded salaries, or cash compensation only? Fully loaded is more honest and produces a larger, less flattering number. If a founder does the selling, is their time in the numerator? It should be, at some rate, or your CAC will look excellent right up until you hire someone and it triples.

Boundary two — customer success. Onboarding is acquisition-adjacent; renewal work is retention. Most teams put the whole function in one bucket and thereby either flatter CAC or inflate it. Split the function by activity if you can, and if you cannot, put it in retention and say so.

Boundary three — brand and content. Spend that produces customers over eighteen months, recognised in the month it was spent, makes a quarter of heavy content investment look like a CAC disaster. Amortising is more economically honest and much harder to defend against accusations of creative accounting. Pick one, and state which.

Boundary four — tooling that serves both. A CRM used by sales and by support is not purely acquisition cost. Allocate by seats, by usage, or by a stated fraction — the specific method matters less than applying it consistently.

The decision rule at every one of these forks: choose the boundary that makes the number *harder* on you. A CAC you have flattered is a number you will act on wrongly, and the cost of that lands later and larger.

Blended, paid and new-logo CAC are three different numbers

Quoting a single CAC hides the composition, and the composition is where the decisions are.

• Blended CAC divides all sales and marketing spend by all new customers, including those who arrived through referral, word of mouth or organic search at no marginal cost. It is the honest headline number and the worst planning number, because it improves whenever your organic channel does and tells you nothing about whether more spend would produce more customers.
• Paid CAC divides paid channel spend by customers attributed to paid channels. It is the number that answers "should we spend more?" and the number most vulnerable to attribution error.
• New-logo CAC excludes expansion revenue from existing accounts. Mixing expansion into the denominator is the most common way a CAC number quietly improves while new-customer acquisition is actually getting harder.

Compute at least blended and paid. The gap between them is a diagnostic in itself: a large gap means your organic and referral engine is carrying the business, which is excellent and also fragile, because it is the part you control least directly.

Payback period is the constraint you actually feel

Bessemer Venture Partners defines the CAC payback period as "a statement in months, of the time to fully payback your sales and marketing investment". It is the more useful of the two headline metrics for most companies, for a simple reason: it is computed from cash you have already spent and cash you will receive on a known schedule. It needs no estimate of how long customers stay.

payback in months = CAC / (monthly revenue per account x gross margin)

Use gross-margin contribution rather than revenue. Paying back ₹50,000 of acquisition cost at ₹6,000 a month of revenue takes eight and a bit months on paper and takes longer in reality, because delivering that revenue costs something.

Bessemer's published guidance is segment-dependent: "for cloud companies selling into SMB-focused accounts, you should target CAC payback <12 months; for mid-market-focused accounts, target CAC payback <18 months; and for enterprise-focused accounts, target <24 months". That variation is not arbitrary — it tracks contract size, sales-cycle length and retention, all of which move together.

The reason payback binds harder than the ratio is cash. A 20-month payback means every new customer is a 20-month loan you are making, funded from somewhere. Growth then consumes cash rather than producing it, and the faster you grow the worse the cash position gets. That is a survivable position with funding and a fatal one without it, which is why the same payback number is a different verdict for different companies.

LTV is the unreliable half of LTV:CAC

Bessemer recommends "investing in customer acquisition when CLTV / CACs are 3x+, and if much under that, continuing to experiment until you have unlocked stronger unit economics". That is sound guidance and it depends entirely on a number most companies cannot compute honestly.

The standard model is:

LTV = monthly revenue per account x gross margin x (1 / monthly churn rate)

The 1 / churn term is the problem. It assumes a constant monthly churn rate, which implies customer lifetimes are exponentially distributed and that a customer in month 30 is exactly as likely to leave as one in month 2. Real retention curves usually flatten — the survivors are structurally stickier than the average — so the model can understate long-run value for good businesses and wildly overstate it for young ones.

It is also arithmetically brittle at the low end, which is where early-stage companies live. Divide by a 2% monthly churn and you get a 50-month lifetime. Divide by 4% and you get 25. A one-in-fifty measurement error in churn halves your lifetime value and halves your ratio, and if your churn estimate comes from four months of data on ninety customers, an error of that size is entirely ordinary.

The rule. Treat LTV:CAC as a directional check that you compute with error bars, and treat payback as the number you actually manage against. If you want to see how sensitive your own figures are, the CAC and LTV calculator implements exactly the model above — gross-margin LTV, the 1 / churn lifespan and the payback division — so moving the churn input a single percentage point shows you the fragility directly.

A worked example, end to end

Every figure below is illustrative. It exists to show the reasoning and the sensitivity, not to describe a benchmark.

A ten-person B2B company reviews the last quarter.

The numerator. Two people spending roughly 60% of their time on acquisition, fully loaded at ₹7,50,000 each for the quarter, so ₹9,00,000 of that is acquisition time. Paid ads ₹1,80,000. Content and design ₹1,20,000. Sales and marketing tooling ₹60,000, of which 75% is allocated to acquisition, so ₹45,000. Total: ₹12,45,000.

The denominator. 18 new customers in the quarter.

Blended CAC = ₹12,45,000 ÷ 18 = ₹69,167.

Now split it. Seven of the 18 came from referrals and organic search with no attributable paid spend. Eleven came from outbound and ads.

Paid CAC = ₹12,45,000 ÷ 11 = ₹1,13,182. Substituting the numbers changes the strategic question completely: the blended figure says acquisition is reasonable, the paid figure says each additional customer you *buy* costs 64% more than the average suggests.

Payback. Average revenue per account is ₹6,000 a month at a 78% gross margin, so ₹4,680 of monthly contribution. Blended payback is ₹69,167 ÷ ₹4,680 = 14.8 months. Paid payback is ₹1,13,182 ÷ ₹4,680 = 24.2 months.

LTV, and its fragility. At 3% monthly churn the model gives a 33.3-month lifetime and an LTV of ₹1,55,844, so blended LTV:CAC is 2.3× and paid is 1.4×. Re-run at 5% churn — a plausible measurement difference on a young cohort — and lifetime falls to 20 months, LTV to ₹93,600, and the two ratios to 1.4× and 0.8×. Nothing about the business changed. One input moved by two percentage points and the paid channel went from marginal to value-destroying.

What this company should conclude. Not "our CAC is too high". The finding is narrower and more actionable: the paid channel has a 24-month payback at a price point that cannot support it, while the referral and organic channel is carrying the blended number. The next move is to fix or shrink paid, and to invest in the channel that is already working — and to measure churn properly before believing any ratio at all.

The levers, ranked by how fast each one moves the number

Ordered by time-to-effect, because the ranking by size is different and less useful when you need the number to move.

1. Stop spending on channels that produce nothing. Immediate. Every quarter has at least one line item that has produced no closed business in two quarters and survives because nobody looked. This requires no experiment and no risk.

2. Remove duplicate tooling. Immediate, and it also removes the per-seat charge that penalises adding a person. Autocloz's free plan covers 5 users and 10 mailboxes with all five outbound channels, booking pages and warmup included, and unlimited seats begin on Growth at ₹1,999 / $24 a month, priced on outbound volume rather than per head — the pricing page sets out where each limit sits. The outbound stack cost calculator is the quickest way to see what your current stack costs against a consolidated one. Be honest about the size of this lever: tooling is usually a single-digit percentage of the numerator, so it is a fast win rather than a large one.

3. Tighten targeting so effort concentrates on fits. Weeks. A narrower profile means fewer sends, fewer meetings and a higher close rate on the meetings you do take, which moves the denominator without moving the numerator.

4. Shorten the sales cycle. One to two quarters. This does not reduce cost per deal directly; it increases deals per unit of rep time, which is the same thing arriving through the denominator. The specific mechanics are in how to shorten your sales cycle.

5. Improve conversion at the weakest step. One to two quarters, and the largest lever in the long run. Find the step with the worst rate relative to its benchmark in your own history, not the step that annoys you most.

6. Raise price, or move upmarket. Two or more quarters. Both change payback directly through the contribution term, and both change everything else about the business at the same time. Real, but not a lever you pull to fix a quarter.

7. Reduce churn. Slowest, and it does not lower CAC at all — it raises LTV and improves the ratio. Worth separating from the CAC levers so nobody expects the acquisition number to move. What actually moves retention is covered in customer retention strategies.

What changes at ten customers a month versus a hundred

At low volume, CAC is mostly noise. Eighteen customers in a quarter means one unusual deal moves the number several percent, and comparing this quarter to last is comparing two small samples. Look at rolling four-quarter figures and resist the urge to explain every movement.

At low volume, attribution is also mostly guesswork. With 18 customers you can ask each one how they found you and get a better answer than any model will produce. Do that; it stops being possible later.

At higher volume, three things change. Attribution becomes a system, with all the error a system introduces — self-reported attribution and click-based attribution will disagree, and neither is authoritative. Segment CAC diverges, so a single blended figure starts hiding a healthy segment subsidising a bad one; split by segment before you split by channel, because segment differences are usually larger. And the payback constraint starts binding on cash rather than on patience, which is the point at which a 20-month payback stops being a metric and starts being a financing decision.

The metric set worth maintaining alongside CAC — and the ones that mislead — is covered in sales metrics and KPIs to track for outbound.

What CAC does not tell you, and what Autocloz does not measure

Some limits, because a metric presented without them gets over-used.

CAC does not tell you whether a customer was worth acquiring. It is an average over a cohort, and averages hide the distribution. Two segments with identical CAC and wildly different retention are the same number and different businesses.

CAC does not distinguish spend that produced this quarter's customers from spend that will produce next year's. Any period-based calculation attributes long-payoff investment to the wrong period, and no boundary choice fully fixes it.

CAC cannot be compared across companies, because it is not standardised and the boundary decisions are almost never disclosed. Treat any competitor's published CAC as marketing.

And Autocloz does not compute CAC for you. It records activity, pipeline and outcomes; the numerator lives in your payroll, your ad accounts and your accounting system, and no CRM sees those. What it can give you honestly is the denominator side — which channels produced which closed business, over what period — and that is the half most teams get wrong.

Autocloz's free plan covers 5 users and 10 mailboxes with all five outbound channels, warmup and DMARC monitoring included, and its AI runs on your own OpenAI, Anthropic or Groq key with no per-lead metering — start free if the tooling line in your numerator is larger than it should be.