Frameworks & methodology: CAC, LTV & ROAS

Your paid team wants another 200,000 next quarter and has a 4.2x ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → deck to justify it. Your CFO wants to know which month the money comes back. Both questions are answered by the same table, and neither is answered by anything inside your ad platform. This lesson builds that table: cohort curves on one axis, contribution margin on the other, with CAC scoped correctly so the ratio at the bottom means something.

The calculation stack

Taking CACCACCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition →, LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → and ROAS as the foundations lesson defines them, the work here is assembling them into one model in a fixed order. Skip a step and the ratio at the end is decoration.

Step one: build the cohort grid. Rows are acquisition months, columns are months since acquisition (0, 1, 2 … 36), cells are cumulative gross profit per acquired customer. Not revenue. Not retention rate. Cumulative gross profit per head, so that any two cohorts can be compared at the same age.

Step two: build the margin stack that fills those cells. Start from revenue per customer, subtract cost of goods or cost of service delivery, then payment processing (2 to 3 percent for most card-based businesses), refunds and chargebacks, variable support cost, and any shipping or fulfilment you eat. What remains is contribution, and contribution is the only number that pays back an acquisition cost. A subscription business with 80 percent gross margingross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition → and 6 percent of revenue lost to failed payments and involuntary churn is running closer to 72 percent.

Step three: separate the observed part of the LTV curve from the extrapolated part. If your oldest usable cohort is 19 months old, you know months 0 to 19 and you are guessing after that. Write the guess down separately: LTV = observed cumulative contribution + modelled tail, with the tail capped at a horizon you state out loud (24 or 36 months is normal) and discounted at your cost of capitalcost of capitalThe blended rate a company pays to finance itself through debt and equity. It sets the minimum return an investment must clear to create value.View full definition →.

Step four: choose the CAC scope that matches the question. Blended CAC answers "is the P&L working". Channel CAC answers "where should the existing budget sit". Marginal CAC answers "what does the next 50,000 buy". They are three different numbers and they diverge most at exactly the moment you are deciding to scale.

Step five: only now compute the ratio and the payback month. A 3:1 LTV to CAC quoted without a horizon and a margin basis is not a benchmark, it is a rumour. 3:1 on 36-month gross-margin LTV and 3:1 on undiscounted lifetime revenue describe two different companies. Target setting itself belongs to the playbook lesson; here the job is making the inputs honest.

Key sub-concepts

1. COHORT CURVES AND THE CENSORING TRAP

Aggregate LTV across your whole base is a weighted average of cohorts you can no longer influence. Cohort curves fix that, but they introduce their own trap: young cohorts have only young data, so any average that mixes ages is biased downward. Always read the grid diagonally, comparing January's month-6 value with June's month-6 value.

The second trap is heterogeneity. Churn is never uniform, so the impatient leave first and observed churn falls with cohort age even when nothing improved. Take a cohort losing 12 percent in month one and 3 percent by month twelve: applying the month-one rate to a simple 1/churn lifespan gives you roughly 8 months of life, while the survivors are plainly worth years. Applying the month-twelve rate gives 33 months and flatters everything. Fit the decay curve to the observed points instead of picking one churn rate, and cap the tail.

1. CONTRIBUTION-MARGIN CAC AND THE CASH TROUGH

Payback is CAC divided by monthly contribution per customer, and it is the number that decides whether growth needs funding. Say CAC is 300 and monthly contribution is 25: payback lands at month 12. Now acquire 1,000 customers a month and grow that intake 10 percent monthly. Each new cohort digs a fresh 300,000 hole before the older cohorts have climbed out, and the cash trough keeps deepening for over a year even though every individual cohort is profitable. This is why two companies with identical LTV to CAC ratios can have completely different survival odds. Payback measures the financing requirement of your growth rate.

1. BLENDED, CHANNEL AND MARGINAL CAC

Spend 100,000 at a 50 paid CAC and you buy 2,000 customers. Push to 150,000 and paid CAC drifts to 60, giving 2,500 customers. The extra 50,000 bought 500 customers, so marginal CAC on that increment is 100, double the average you are reporting. Auctions, audience saturation and diminishing frequency response mean the marginal curve always sits above the average curve, and the gap widens as you scale. Budget decisions are marginal decisions, so measure the increment, not the average. Meanwhile blended CAC drifting upward while paid CAC holds flat is usually organic dilution, not paid inefficiency: you are buying customers who used to arrive free.

1. ROAS FLOORS DERIVED FROM MARGIN

Break-even ROAS is 1 divided by contribution margin. At 30 percent contribution the floor is 3.33x before you have paid a single salary. At 70 percent it is 1.43x. Any team running one ROAS target across a portfolio with mixed margins is subsidising its worst products with its best. Set floors per channel role too, and treat retargetingretargetingShowing ads to users who have previously visited your site or interacted with your brand, to bring them back and drive conversion.View full definition → ROAS with suspicion: a 12x retargeting return usually reflects credit reallocation rather than demand creationdemand creationCreating and stimulating demand for your offer, often upstream of the buying process to generate interest and awareness before prospects are ready to buy.View full definition →, which is why holdout tests exist.

Real-world cases

CASE 1: Spotify and the margin denominator

Spotify pays the bulk of its revenue to rights holders, which has kept gross margin in the mid-20s percent for most of its listed life, improving to around 30 percent in 2024. Premium ARPU sits in the region of 4 to 5 euros a month, so contribution per subscriber is well under 2 euros a month. Run the payback arithmetic: a 20 euro acquisition cost needs more than a year of survival to clear, and that is before discounting. The same 20 euro CAC in an 80 percent margin business clears in a quarter. Spotify also raised US Premium pricing in 2023 and again in 2024, which lifts the back half of every surviving cohort's curve. Any LTV model that freezes ARPU at the acquisition-month value will understate old cohorts and misprice new ones.

CASE 2: Shopify and expansion inside the curve

Shopify's revenue is split between subscription plans and merchant solutions tied to the GMV its merchants process, with merchant solutions now the larger share. That changes the shape of the LTV curve: surviving merchants grow, so cohort contribution per head can rise in later months even as the number of merchants falls. Trial and first-year merchant attrition is heavy, so months 0 to 6 look poor while the tail carries most of the value. Model that with a flat retention assumption and you will conclude the business does not work. (Shopify sells the commerce platform and publishes its own guidance on these metrics, so read its benchmark posts as marketing material with useful arithmetic inside.)

CASE 3: Wise and the organic cannibalisation problem

Wise has said for years that roughly two-thirds of its new customers arrive through word of mouth, and it has around 12.8 million active customers moving well over 100 billion pounds a year at a take rate under 1 percent. Two consequences for the model. First, revenue per customer is small and frequency does the work, so LTV is a volume curve, not a subscription curve. Second, Wise's strategy of repeatedly cutting prices reduces revenue per transfer while lifting volume and referral, which any static ARPU-based LTV formula will score as value destruction. Third, and this is the one that catches teams: when organic supplies two-thirds of intake, paid campaigns partly buy people who were coming anyway, so measured paid CAC understates true incremental CAC and marginal CAC rises steeply.

CMO action items

• Ship the cohort grid before you ship the dashboard: acquisition month by age, cumulative contribution per customer, 36 columns. Everything else derives from it. If your data team cannot produce it in two weeks, that is the first problem to solve.
• Agree the margin stack line by line with finance and freeze it in writing. Which costs sit above the contribution line decides your ROAS floor, and unilateral marketing definitions of margin do not survive a board meeting.
• Report marginal CAC alongside blended CAC in every budget request, with the spend increment and customer increment shown.
• Publish your LTV horizon and discount rate on the same slide as the ratio. An undated ratio invites everyone to assume their own number.

Common mistakes that kill results

Mistake 1: revenue ROAS instead of contribution ROAS

A fashion brand at 4x ROAS looks fine. With 60 percent cost of goods and a 25 percent return rate, the contribution behind that 4x is close to break-even. Recompute every floor in contribution terms once, and most channel rankings reshuffle.

Mistake 2: one average LTV for the whole base

Averages hide the spread. When a minority of customers carries most of the value, a universal CAC target overpays for the low-value tail and starves the segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition → worth chasing. Segment LTV by acquisition channel, first product, geography and initial order size before any CAC target is set, then let the CAC ceiling vary by segment.

Mistake 3: ignoring the time value of money

An LTV of 500 spread over five years is not 500 today. At a 10 percent annual cost of capital that stream is worth roughly 380 in present value, and the gap widens with every rate rise. Discount the tail, or your payback targets will be systematically too generous.

Mistake 4: counting trials and refunds as acquisitions

Anyone who refunds inside the guarantee window, or never converts from trial, belongs in neither the numerator nor the denominator. Leaving them in inflates the customer count, deflates CAC and produces a curve that starts above where any real cohort starts.