How to Calculate Unit Economics: CAC, LTV, Payback

All right, here is the direct answer. CAC is everything you spent on sales and marketing in a period, divided by the new customers that spending produced. LTV is average revenue per account, times gross margin, divided by churn. CAC payback is CAC divided by the monthly gross profit one customer brings in. Then compare them.

Unit economics is one question asked three ways. Does a customer give back more than it cost to win them, by how much, and how fast. CAC is the cost, LTV is the amount, payback is the clock. The familiar rules of thumb, LTV at least three times CAC and payback inside twelve to eighteen months, are those three numbers set against each other, and they are the easiest numbers in a board deck to get wrong quietly.

Here is the reframe I would offer. These are not accounting numbers. They are management estimates wearing accounting clothes, and every input is a choice somebody made: what counts as a sales cost, which quarter's spend matches which quarter's customers, whether gross margin is in the formula, how long a customer lives. The closest thing to a rule is the SEC's guidance on metrics like these, which asks public companies for "A clear definition of the metric and how it is calculated," why it is useful, and how management uses it.¹ If you cannot write those three sentences about your CAC, you have a number, not a metric.

So let me walk you through it in the order I build it. CAC, including the costs people leave out. Why a blended CAC hides something. LTV and the two mistakes that turn it into fiction. The 3 to 1 rule and what it cannot tell you. Payback, the one I trust most. A cautionary tale from a public filing. Then the order I run them in.

CAC is simple right up until somebody asks what you left out

The formula is one line. Total sales and marketing spend for a period, divided by new customers won in that period. The trouble lives in the word total. Andreessen Horowitz put it plainly: "Customer acquisition cost or CAC should be the full cost of acquiring users, stated on a per user basis."² Full means salaries for everyone in sales and marketing, plus tools, agencies, events, commissions, referral fees, and the credits and discounts you handed out to close. Brian Balfour, who ran growth at HubSpot before founding Reforge, names the first mistake he sees: "You need to include the salaries of all people working on marketing and sales."³ That is fully loaded CAC, and it is the only version I will sign off on.

The second problem is timing. Money spent in one quarter closes deals in the next, so matching this quarter's spend to this quarter's customers can flip a decision. Balfour's worked example shows a March CAC of $148 against a $125 target when spend and customers sit in the same month, and $84 once you allow for the two month lag from lead to customer.³ Same business, same money, opposite conclusion about whether to cut the channels.

David Skok's convention, in the definitions behind SaaS Metrics 2.0, is to set the prior quarter's spend against this quarter's new business, on the thought that there is a delay between the money going out and the deal closing, unless your cycle is short enough to use the same quarter.⁴ CJ Gustafson lags the spend to the sales cycle you are measuring, so a 180 day enterprise motion gets a two quarter lag and self serve gets none.⁶ Pick one, write it down, and never change it quietly. Brad Coffey of HubSpot, in the same article, named the real enemy as the pressure to spin, and his two rules were to count every cost and stay consistent.⁷

Investors usually look at the same thing upside down, as cost per dollar of new recurring revenue. Dave Kellogg calls it the CAC ratio and reads roughly 1 as healthy,⁸ and Scale Venture Partners turned it into the magic number in 2010 after Rory O'Driscoll decided that for subscription businesses "Classic sales and marketing efficiency metrics are utterly misleading."⁹ Benchmarkit's 2026 survey of 342 software companies gives the current state of play: "Median Blended CAC Ratio is $1.30, meaning $1.30 of S&M expenses per dollar of new ARR," with the median magic number at 1.37, above 1.0 for the first time in four years.¹⁰

Blended CAC hides the truth, and here is how to see it

A blended CAC divides all acquisition spend by all new customers, including the ones who arrived on their own. Andreessen separates it from paid CAC, spend divided by customers won through paid marketing, and says why: blended is not wrong, but it cannot tell you whether your paid campaigns are profitable.² Balfour is blunter: "For internal operations and decision-making however, average CAC is almost always useless."¹¹ The organic customers subsidize the paid ones, and the blend looks fine right up until the organic well runs dry.

The cleanest example I know is HubSpot's own, told by Coffey inside Skok's article. In their very small business segment, HubSpot measured an LTV to CAC of 1.5 selling direct and 5 selling through channel partners. One number said stop and one said go. Within twelve months they went from 12 direct reps and 4 channel reps to 2 direct and 25 channel.⁷ A blended CAC across both motions would have been a mediocre number nobody acted on. Segmented, it rebuilt the sales organization.

So cut it every way that changes a decision: by channel, by segment, by geography, self serve against sales assisted. Then watch the marginal number rather than the average, because acquisition costs climb with scale: Andreessen's rule of thumb runs a dollar a user for the first thousand and five to ten for the next hundred thousand.² The average always looks better than the last customer you bought.

LTV, and the two mistakes that turn it into fiction

Lifetime value is the gross profit a customer generates over the time they stay. The working formula is average revenue per account, times gross margin, divided by churn, where one divided by churn is the expected lifetime. Skok's definitions put the margin piece plainly: "To truly get an accurate picture of LTV, you should take into consideration Gross Margin."⁵ Three percent monthly churn implies a 33 month lifetime, and 20 percent annual churn implies five years.

Christoph Janz of Point Nine states the first mistake without hedging: an LTV built on revenue instead of gross profit makes no sense.¹² Two businesses at 80 and 50 percent margin can show identical revenue LTVs and be completely different companies. For an AI product carrying inference cost inside cost of revenue, that gap is the whole story.

The second mistake is the lifetime itself. One divided by churn assumes churn is constant and forever, and at low churn the number runs away from you. Skok said it when he replaced his own formula: with long lifetimes and negative churn the old LTV "can become infinite," and the CFOs he talked to already knew it.¹³ His fix was to discount the cash flows: "We recommend using a 10% discount rate, but this may differ depending on your own cost of capital."¹³

Peter Fader and Bruce Hardie showed that company level retention looks steady only because it blends young cohorts with old ones: "The relatively constant retention rates observed in the company-reported summaries are in fact the result of aggregation across different cohorts of customers."¹⁴ That is why cohorts come before formulas. Janz again: "The best way to approximate LTV is to take a close look at your cohorts."¹²

In July 2026 ChartMogul tested the basic formula at scale, 35,512 cohort observations across 3,331 companies, comparing the LTV predicted at signup against the revenue those cohorts actually produced. The median cohort came in 2.4 percent under its prediction at twelve months, which sounds fine, except that 28.3 percent of cohorts missed by more than half in one direction or the other.¹⁵ Their conclusion: "LTV prediction errors are not random. They are systematic, making the metric reliable in some situations and consistently misleading in others."¹⁵

Bill Gurley wrote the warning label in 2012: "The fundamental reason that it is so amazingly dangerous and seductive is its simplicity and certainty."¹⁶ His deepest objection is that the inputs move together: raise price and churn rises, spend more on acquisition and cost per customer rises, improve churn with service and cost to serve rises. Use LTV to compare channels and cohorts, which is what he says it is for, not to justify spending money you have not earned yet.

The 3 to 1 rule, where it came from, and what it cannot tell you

The rule that LTV should be at least three times CAC is not a law of nature. Skok introduced it as a viability guideline and then checked it against the companies he worked with: "The best SaaS businesses have a LTV to CAC ratio that is higher than 3, sometimes as high as 7 or 8."⁷ Bessemer's growth team frames it from the investor's chair: "As investors, we consider 3x+ to be solid and 5x+ best in class."¹⁷

The current data sits above the rule. The 2026 Aleph and Benchmarkit benchmarks, from the 146 companies in their 342 company sample that reported it: "The median B2B SaaS company has a CLTV:CAC ratio of 4.1x," with the top quartile at 7.8 and the bottom quartile at 1.1, where lifetime value barely covers acquisition.¹⁸ A very high number is not automatically good, either. HubSpot's guidance notes that five to one can mean you are underspending on sales and marketing,¹⁹ which is exactly the conversation I want with a proud, profitable founder.

And early on, the ratio is mostly imagination. Tomasz Tunguz calls it false confidence for young companies, because the L in LTV is a forecast you cannot make yet: "At 10% unit churn, three years from now, 73% of customers will still be paying, adding to their LTV."²⁰ Nobody two years into selling knows what happens after that. He points to payback instead, because within 14 to 18 months most startups have real payback data.

Payback is the number I trust most, because it is a risk number

CAC payback is how many months of gross profit it takes a new customer to repay what you spent to win them. Bessemer named it one of five metrics that matter for cloud companies in 2012: "The CAC payback period is a statement in months, of the time to fully payback your sales and marketing investment."²¹ The gross margin adjustment is not optional, because delivery cost never becomes profit, and the public market versions agree: Jamin Ball's Clouded Judgement runs prior quarter sales and marketing over net new ARR times gross margin, times twelve.²²

Kellogg has the sharpest line on what the number is for: "They forget payback metrics are risk metrics, not return metrics."²³ It tells you how long your cash is exposed, not how much you will make. A customer paying $150 a month at 70 percent margin and 3 percent monthly churn against a $3,500 CAC shows a 33 month formula payback and in reality never pays back, because churn eats it first.²³ SaaS Capital, a lender to software companies, explains why: "the faster the payback, the faster profits can be recycled back into acquiring new customers."²⁴

Now the benchmarks, with their samples attached. Bessemer's targets by segment: "For cloud companies selling into SMB-focused accounts, you should target CAC payback <12 months;" under 18 for mid market and under 24 for enterprise, with a 15 month average across their portfolio between $1 million and $10 million of ARR.²⁵ The 2026 Aleph and Benchmarkit data, 198 companies reporting: "The median B2B SaaS company recovers its customer acquisition cost in 16 months," down from 18 the year before, with the top quartile at 6 months or fewer and the bottom quartile at 24 or more.²⁶

Skok's original guideline was under twelve months, written in 2011, and his definitions page now calls around 20 common and anything over 24 a signal to fix something.⁵ So twelve to fifteen months is a fair target for a self serve or SMB motion and a stretch for enterprise. One word of caution from Gustafson: he puts the odds that a self reported payback is understated at nearly 100 percent, because overhead and stock compensation get left out of sales and marketing and out of cost of revenue.⁶ When I rebuild one from the general ledger, the number usually gets worse before it gets honest.

The cautionary tale I would rather you hear from me: Blue Apron

Blue Apron is not software, but it is a subscription business whose unit economics were argued in public, in real filings. Its June 2017 prospectus defined Cost per Customer as cumulative marketing spend over cumulative customers, $94 from 2014 through the first quarter of 2017, and set it against cumulative net revenue per customer of $410 at six months rising to $939 at 36 months, before cost of goods sold.²⁷ On contribution, the filing said "our net contribution per Customer for the six month period after such Customer's first order was $115," or 1.2 times the $94.²⁷ Six month revenue per customer for the 2016 cohort had slipped to $387 from $451, which the company put down to promotional discounts.²⁷

Daniel McCarthy, then at Emory, rebuilt the customer economics from those disclosures the day before pricing: "We estimate that CAC in Q1 2017 is $147."²⁸ He estimated that 72 percent of customers churned by month six, that a new customer needed about $565 of net revenue to break even at a 26 percent variable contribution margin, and that roughly 70 percent of recent customers would never get there.²⁸ His $147 against the company's $94 is the gap between a marginal, cohort based number and a cumulative blend of three years of cheaper customers.

The IPO priced at $10 a share on June 28, 2017, cut from an original range of $15 to $17.²⁹ By the 2018 annual report, net revenue had fallen 24 percent to $667.6 million, customers had gone from 1,036 thousand in the first quarter of 2017 to 557 thousand at the end of 2018, the stock had closed at $3.35 the previous June, and the filing carried a risk factor about the exchange's one dollar minimum price.³⁰

Nothing in the prospectus was false. Every number in it was blended, cumulative, or before cost of goods, and the person who computed the marginal, cohort based, after contribution version was outside the building. I have sat on the sell side of two private equity exits, and a quality of earnings team computes them the outside way, from your source data, whether or not your deck did. Better that you get there first.

The order I actually run these in

When somebody sends me a metrics page cold, I start with the inputs, in this order, and I write the definition of each down before I calculate anything, because the definition is what a buyer will test first.

1. Fully loaded CAC, by segment and channel, with the spend lagged to the sales cycle. Every salary, tool, agency, commission, credit and discount. Blended for the board slide, segmented for every decision.
2. Gross margin adjusted CAC payback, in months, booked and cash. CAC divided by monthly revenue per customer times gross margin. This is the number I trust first because it needs the least forecasting.
3. Cohort retention curves before any lifetime. Logo and revenue retention by signup month, out as far as the data goes. If the curve is still falling, you do not know the lifetime yet, and neither does your LTV.
4. LTV on gross profit, capped or discounted. Average revenue per account times gross margin divided by churn, then cap the lifetime at what the cohorts have actually shown, or discount the cash flows the way Skok does.
5. LTV to CAC, read next to payback and net revenue retention. Above 3 is the guideline, the 2026 median is about 4, and a number above 5 is a question about underinvestment, not a trophy.

Two closing notes. High Alpha's 2025 benchmarks, from more than 800 respondents, found that retention and acquisition efficiency together predict performance, and the relationship is lopsided: "Even modest increases in NRR can offset higher CAC, but the inverse rarely holds true."³¹ So if you can only fix one input, fix retention. And if you are building on AI with real inference cost, ICONIQ suggests a gross margin adjusted magic number, because strong sales efficiency on a compressed margin flatters a business that is not efficient at all.³² Margin first, then unit economics, then the story.