CAC, LTV and payback
Compute CAC, LTV and the payback month from one month of spend, and size the cash need that an LTV/CAC of 1.5 hides.
On this page
Three numbers say whether a company can afford the way it acquires customers. CAC is the spend per new customer, LTV is the margin a customer returns over their life, and payback is the month in which cumulative margin first covers the CAC. The first two make a ratio that fits on a board slide. The third says how long the cash is out, and for the invoicing app it says something the ratio cannot.
CAC: spend over the customers it bought
The funnel lesson followed one month of paid visits to 96 paying customers. That month cost $50,000 of paid search and paid social:
CAC = paid spend ÷ customers that spend produced
= $50,000 ÷ 96 = $521
Two choices hide in that line. The numerator is media only and the denominator is paid customers only; the 64 customers who arrived without a paid click, and the salaries that run both channels, are the subject of blended versus paid CAC. And the 96 are the customers January's spend produced, not the ones who happened to start paying in January. A visitor in this funnel pays only after three months of free use, so dividing this month's spend by this month's new payers mixes cohorts. With flat spend the error is small. Just after a cut in spend it flatters: this month's payers were bought by the bigger months before, and the shortcut reports a CAC lower than the one now being bought.
LTV: margin, not revenue
The paid plan is $50 a month at 80% gross margin, so a customer returns $40 a month. With 5% of customers leaving each month, the average customer stays 1 ÷ 0.05 = 20 months:
LTV = monthly margin ÷ monthly churn = $40 ÷ 0.05 = $800
Revenue would say $50 × 20 = $1,000, which includes $200 the company never keeps. Margin is the right base because acquisition is paid from what is left after serving the customer. That matters more for a product whose features call a language model, because each call is a cost of serving; where the money goes shows how it scales with every request. If model calls cost $5 per customer a month, margin falls to $35 and LTV to $700 before marketing has changed anything.
The ratio is 800 ÷ 521 = 1.54. A customer returns about one and a half times what it cost, so every customer is profitable over their life. That sentence is true, and it is where the trouble starts.
Payback, the flattering way
The quick payback divides CAC by monthly margin:
contracted payback = $521 ÷ $40 = 13.03, so month 14
This counts the margin the contract would pay if every customer stayed, which is why it is called contracted margin. It holds only with zero churn. At 5% a month, 0.95¹³ ≈ 51% of the cohort is still paying in month 14, so the cohort has collected far less than 14 × $40 for each customer it started with.
Payback on expected margin
The honest version cumulates what the cohort is expected to pay. In month 1 everyone pays $40, in month 2 95% do, in month 3 0.95² do, and so on. Summed to month n:
cumulative expected margin = 40 × (1 + 0.95 + … + 0.95^(n−1)) = 800 × (1 − 0.95^n)
800 × (1 − 0.95^n) ≥ 521 → 0.95^n ≤ 0.349 → n ≥ 20.5, so month 21
That is seven months later than the quick version, and the curve can never pass $800, however long it runs. The same formula gives month 11 for the $333 customer from module 0's paid-social chain in the arithmetic a marketer carries, and the course's 12-month payback target sets the CAC ceiling at 800 × (1 − 0.95¹²) ≈ $368. The invoicing app's paid CAC is $153 over it. This is the definition the module's code exercise, payback month from a cohort, implements, and the one the rest of the course uses.
The cash the ratio hides
A ratio describes one customer. A company runs the program every month. Spend $50,000 a month, buy 96 customers each time, and in month t the cohorts bought so far return 96 × 800 × (1 − 0.95ᵗ), which is 76,800 × (1 − 0.95ᵗ):
| Month | Margin returned | Net that month | Cumulative |
|---|---|---|---|
| 1 | $3,840 | −$46,160 | −$46,160 |
| 12 | $35,300 | −$14,700 | −$349,000 |
| 20 | $49,300 | −$700 | −$400,000 |
| 21 | $50,600 | +$600 | −$399,000 |
| 51 | $71,200 | +$21,200 | +$14,000 |
Every month loses money until month 21, and the cumulative position bottoms at about −$400,000 in month 20. The program earns that back in month 51. An LTV/CAC of 1.5 is a promise that the money returns; payback says the company must find $400,000 and wait more than four years to collect it. At module 0's $333 CAC the same $50,000 buys 150 customers a month, the low is about −$215,000 in month 10, and the program is whole by month 23.
The table flatters in one way, too: it starts each cohort's margin in the month the cohort is bought. A visitor in this funnel pays only after three months free, so three months of spend, $150,000, go out before the first margin arrives. Counted that way the low is nearer −$550,000, in month 23, and the program is whole in month 61. The table is the least cash the program needs.
What moves payback
Three inputs set the month: CAC, margin and churn. Moved one at a time from the running numbers:
- CAC down to the $368 ceiling: month 12, by construction. It is a 29% cut, usually made by stopping the worst third of spend rather than improving all of it.
- Churn from 5% to 4%: LTV rises to $1,000, a quarter more, and payback improves only to month 19.
- Margin from $40 to $35, the model-cost case above: month 27, six months later, from a $5 cost per customer.
Payback is set by the early months, so at these numbers it responds to CAC and margin first and to churn last; churn mostly changes the months after payback. Where churn is higher the balance shifts, and the code exercise shows a retention change worth four months at 10% churn. Either way a retention project is valued in months of cash, not in LTV alone.
In practice
Show the board payback on expected margin, with LTV/CAC beside it and never alone: "Paid CAC is $521 and pays back in month 21 against a 12-month target. At this spend the program needs at least $400,000 of cash before it turns, nearer $550,000 with the three free months, and the lever that closes most of the gap is cutting CAC towards $368." A board can approve that, or decide the cash is better spent elsewhere. It can do neither with "LTV/CAC is 1.5".
Ask finance to compute payback per cohort from the cohort table once six months of data exist, instead of from a single churn rate, and to count the cost of serving, model calls included, in margin.
The mistake that costs money is approving spend on the LTV/CAC ratio. At 1.5 every customer is profitable, and the program that acquires them needs at least $400,000 of cash and more than four years to prove it.
Check yourself
1. A partner channel offers the same customer at a $450 CAC with no minimum. Is it inside the 12-month payback target, and do you take it?
800 × (1 − 0.95ⁿ) ≥ 450 needs 0.95ⁿ ≤ 0.44, which first holds at n = 17. Outside the target by five months, but $71 cheaper than paid and four months faster. Take it in place of the most expensive paid spend, not on top of it, so the cash gap does not grow.
2. Raising the price from $50 to $60 lifts margin to $48 a month, but you expect churn to rise from 5% to 6%. LTV is unchanged at 48 ÷ 0.06 = $800. Does the price rise help?
Yes, if cash is the constraint. Cumulative margin becomes 800 × (1 − 0.94ⁿ), which reaches $521 at n = 18 instead of 21, because more of the same $800 arrives in the early months. LTV/CAC is 1.54 either way, so the ratio cannot see the difference; payback can.
3. The founder proposes doubling paid spend to $100,000 a month at the same $521 CAC. Estimate the cash the program needs before it turns.
Twice the customers at the same CAC doubles every row of the table: a low of about −$800,000 around month 20, whole again in month 51, and nearer −$1.1 million once the three free months are counted. The same CAC is also optimistic, because the second $50,000 reaches many of the people the first already reached. Approve it only with that cash in hand, and measure the CAC of the second $50,000 on its own.