Unit Economics in Plain English: CAC, LTV, Payback and Margin

In 2015, four people at the venture firm Andreessen Horowitz published “16 Startup Metrics,” a guide to the numbers investors ask founders about. One line ends with an exclamation mark: “GMV does not equal revenue!”

GMV is the total value of everything sold through a marketplace. Revenue is the slice the marketplace keeps. The authors felt they had to shout, because founders kept presenting the big number as if it were theirs.

That mistake is a small version of a bigger problem. A business can look busy and growing, and still lose money on every customer. Selling more doesn’t fix that. It makes it worse.

Unit economics is how you find out, before the bank account tells you. It asks one simple question: when one customer arrives, pays you and eventually leaves, did you end up with more money than you spent to get them?

Below is every number you need, each with a definition, a formula and a worked number. Then three fictional businesses go through the whole thing.

Start with the “unit”

The unit is the one thing you count: usually a customer, sometimes an order, booking or project.

Pick the one that matches how money comes in: a paying account for software, a regular customer (or a job) for a service, a buyer or an order for a marketplace, a booking for a rental.

Which you pick matters less than staying consistent. Acquisition cost per signup against revenue per paying customer mixes two different units, and the result will lie to you.

The building blocks

All numbers are per month. Keep everything in the same period: mixing monthly churn with yearly revenue makes you wrong by a factor of twelve.

Revenue per unit (ARPU)

What one customer pays you, on average, in a period. ARPU stands for “average revenue per user.” Some people say ARPA, “per account.” Same idea.

ARPU = total revenue in the month / number of paying customers that month

If 500 customers paid you $15,000 this month, ARPU is $15,000 / 500 = $30.

Gross margin

The share of each dollar left after paying the direct cost of delivering what you sold. Those direct costs are called cost of goods sold, or COGS.

Gross margin = (revenue - cost of goods sold) / revenue

If a customer pays $30 and delivering the product to them costs $6, gross margin is ($30 - $6) / $30 = 80%. The $24 left over is gross profit.

What counts as cost of goods depends on the kind of business:

The rule of thumb: if the cost goes up when you serve one more customer, it probably belongs in cost of goods. If not, it’s a fixed cost.

Contribution margin

What’s left from each customer after every cost that grows with that customer, before your fixed costs.

Contribution margin = revenue - all variable costs

It’s a stricter gross margin: it also takes out variable costs outside cost of goods, such as sales commissions or per-order packaging.

If a cleaning customer pays $120 a month, the cleaner and supplies cost $78, and travel and payment fees add another $6, then gross profit is $42 and contribution is $36. That $36 is what the customer “contributes” toward rent, your salary and profit.

Churn and retention

Churn is the share of customers who leave in a period. Retention is the share who stay.

Monthly churn = customers lost this month / customers at the start of the month Retention = 1 - churn

If you start the month with 500 customers and 20 cancel, churn is 20 / 500 = 4%, and retention is 96%.

Small founders skip churn most often, because it’s uncomfortable. It also changes everything else. Watch out for net churn, which subtracts revenue from upgrades and can look tiny while customers walk out the door. The a16z guide notes it understates your losses. Track customer churn first.

Customer lifetime value (LTV)

The gross profit you expect from one customer over the whole time they stay.

LTV = ARPU x gross margin / monthly churn

With the numbers above: $30 x 80% / 4% = $24 / 0.04 = $600.

Why: if 4% leave each month, the average customer stays 1 / 0.04 = 25 months, and 25 x $24 = $600.

This formula makes big assumptions, and they all lean toward a bigger number:

1. Churn never changes. In real life it moves with competitors, prices and seasons.
2. ARPU never changes. No downgrades, no discounts ending.
3. A dollar in month 40 is worth a dollar today. No allowance for risk or for needing cash now.
4. Your churn number is accurate. After three months, you don’t know your churn. You know your first three months.

The last one causes the biggest errors. At 1% churn the formula says $2,400, because it assumes customers stay 100 months, over eight years. An eight-month-old company has no evidence for that.

A safer habit is a capped LTV: gross profit over the first 24 months, counting customers who leave along the way. At 4% churn that’s about $375, not $600. The a16z guide also suggests 12- and 24-month LTV when data is thin, and says LTV should be based on profit, not revenue.

Customer acquisition cost (CAC)

What you spend, on average, to win one new paying customer.

CAC = sales and marketing spend in the period / new paying customers in the period

Spend $12,000 on ads and get 30 new customers, and CAC is $12,000 / 30 = $400.

• Paid CAC counts only customers who came from what you paid for.
• Blended CAC divides the same spend by all new customers, including word of mouth and search.

If those 30 paid customers came alongside 30 who found you for free, blended CAC is $12,000 / 60 = $200. Half the cost on paper.

The a16z authors say paid CAC matters more when judging whether the business works, and I agree. Free customers don’t scale on command. Paid ones do. And paid CAC tends to rise as the cheapest audiences run out. Count everything: ads, tools, freelancers, first-order discounts, and a fair share of your own sales time.

LTV:CAC ratio

How many dollars of lifetime gross profit you get for each dollar spent to win a customer.

LTV:CAC = LTV / CAC

$600 / $200 = 3.0.

You’ll often hear that this should be at least 3 to 1. The best-known written version comes from David Skok, a venture capitalist and former founder, in his widely read “SaaS Metrics 2.0” guide on forEntrepreneurs.com. He suggests the ratio should be higher than 3 for a healthy subscription software business. He’s also clear that this guideline assumes a simple LTV calculation and high gross margins, around 80% or more.

So treat 3:1 as a rule of thumb from venture investors looking at software companies, not a law. It’s a smell test, and a ratio built on a made-up LTV is a made-up ratio. Below 1, each customer loses money. Far above 5 can mean you’re under-spending on growth, or your LTV is too optimistic.

CAC payback period

How many months of gross profit it takes to earn back what you spent to win a customer.

CAC payback (months) = CAC / (ARPU x gross margin)

$200 / $24 = 8.3 months.

For small companies I like this more than LTV:CAC, because it doesn’t guess the future. Until a customer pays back, you’re lending them money, and if you’re self-funded, the length of that loan limits how fast you can grow.

Skok’s original guideline was under 12 months, though he notes that’s now rare in software and some enterprise businesses work at around 20. Self-funded, shorter is safer. And if payback is longer than the average lifetime, most customers never pay you back.

How marketplaces are different

A marketplace connects buyers and sellers and takes a cut. That creates two traps.

GMV vs revenue. Gross merchandise value (GMV) is the total amount buyers paid sellers through your platform. Your revenue is only your cut of it.

Revenue = GMV x take rate

Take rate is the percentage you keep. Sell $100,000 of game accounts at a 10% take rate and your revenue is $10,000. But card processing is charged on the full $100,000, which eats into your thin slice fast.

Two sides to acquire. Buyers and sellers both cost money to attract. A simple approach is to model buyer economics first and track seller acquisition as its own cost line.

Three fictional businesses, worked through

These are made-up companies with made-up numbers, built to show the method. None of them is a real business of mine or anyone else’s.

Example 1: “Ledgerly,” a small SaaS

A bookkeeping tool for freelancers at $30 a month.

What it says: Blended, Ledgerly hits the 3:1 rule exactly and pays back in about eight months. Paid, it’s at 1.5:1 with a payback of nearly 17 months. That’s two thirds of the 25-month average lifetime, and a paid customer’s realistic 24-month LTV ($375) is less than the $400 it cost to win them. Free customers are carrying the paid ones, and doubling the ad budget mostly buys more 1.5:1 customers.

Example 2: “Spotless,” a home-cleaning service

Regular customers get two cleans a month at $60 each.

What it says: A 35% gross margin looks weak next to Ledgerly’s 80%, yet Spotless pays back in about four months. Low margin is fine when customers are cheap to win and stay a while. The risk is elsewhere: $36 a month per customer must cover the owner’s time, insurance and quiet months. Each customer is worth having, but whether the whole business profits depends on fixed costs, which is what Default Alive or Default Dead? is about.

Example 3: “LootSwap,” a game-account marketplace

Players buy and sell game accounts. LootSwap holds the payment until the buyer confirms, and takes 10%.

What it says: The pitch-deck number is $312.50 of lifetime GMV per buyer. The real number is $15.63 of gross profit. Against a $25 paid CAC, every buyer from ads loses about $9. The 20-month payback is longer than the 12.5-month average lifetime, so most paid buyers never pay back at all.

The fixes are rows in the table: a higher take rate, lower fraud losses, buyers who return more often, or cheaper acquisition through search and community. Each lever moves one specific number.

The money path of one customer

Here are Ledgerly’s blended numbers as a picture: $200 out on day zero, $24 back each month, zero crossed a little after month 8.

This customer stays all 24 months and brings $576 of gross profit. Averaged over everyone, including those who leave, it’s $375.

A tiny Python sheet you can reuse

All the arithmetic above comes from this. Change the inputs to your own numbers.

def unit_economics(arpu, gross_margin, monthly_churn, cac, months_cap=24):
    gp = arpu * gross_margin              # gross profit per customer per month
    lifetime = 1 / monthly_churn          # average months a customer stays
    ltv = gp / monthly_churn              # simple LTV
    capped, alive = 0.0, 1.0              # LTV over the first N months
    for _ in range(months_cap):
        capped += alive * gp
        alive *= 1 - monthly_churn
    return {
        "gross_profit_per_month": round(gp, 2),
        "lifetime_months": round(lifetime, 1),
        "ltv_simple": round(ltv, 2),
        f"ltv_{months_cap}m": round(capped, 2),
        "ltv_to_cac": round(ltv / cac, 2),
        "payback_months": round(cac / gp, 1),
    }

# Ledgerly (SaaS), blended and paid CAC
print(unit_economics(30, 0.80, 0.04, 200))
print(unit_economics(30, 0.80, 0.04, 400))
# Spotless (service), gross margin 35%
print(unit_economics(120, 0.35, 0.06, 150))
# LootSwap (marketplace), per buyer: $2.50 revenue/month, 50% margin, paid CAC
print(unit_economics(2.50, 0.50, 0.08, 25))

The first line prints a gross profit of 24.0, a 25-month lifetime, a simple LTV of 600.0, a 24-month LTV of 374.75, a ratio of 3.0 and a payback of 8.3 months. The LootSwap line prints a ratio of 0.62 and a payback of 20.0 months.

What good and bad look like

Rough guides, mostly from subscription software. Adjust for your business.

Direction often matters more than level: a 14-month payback that shrinks every quarter beats 9 months that keeps growing.

Common mistakes

1. Counting revenue as profit. LTV built on revenue can be several times too big. For LootSwap it would be $312.50 of GMV or $31.25 of revenue instead of $15.63 of gross profit.

2. Ignoring churn. “$30 a month” means nothing without “for this long.” No churn number, no LTV.

3. Letting blended CAC hide paid CAC. Ledgerly looks like 3:1. Its paid channel is 1.5:1. A plan to “spend more on ads” built on the blended number is built on the wrong number.

4. LTV from too little data. Three good months, often from your friendliest early customers, don’t predict the next thirty. Use a 12- or 24-month LTV until you have history.

5. Mixing periods, units and hidden costs. Monthly churn with annual ARPU. CAC per signup with LTV per paying customer. Forgetting refunds, chargebacks and first-order discounts. Each looks small and moves the answer a lot.

6. Thinking good unit economics means a profitable business. It means each customer is worth having. You still need enough of them to cover fixed costs.

A one-page unit-economics sheet

One sheet, one row per month, these columns. It fits into the setup in The One Google Sheet That Runs a Small Business.

Use the 3-month average churn for LTV, not one lucky month, and fill in the notes. Six months later, they’re the only way to know why a number moved. In my view, that’s the real value of the sheet: it turns guesses about churn and acquisition cost into numbers.

Key ideas

• Unit economics asks one question: does one customer bring in more gross profit than they cost to win?
• Use gross profit, not revenue, and for marketplaces remember that GMV is not revenue. Revenue is GMV times your take rate.
• Churn drives everything. The simple LTV formula (ARPU x gross margin / monthly churn) assumes it never changes, so it often overstates. Check a 24-month LTV too.
• Paid CAC is the number you’ll live with when you try to grow. Blended CAC can hide a channel that loses money.
• 3:1 LTV:CAC is a rule of thumb from venture investors looking at software companies, not a law.
• Payback months is the most practical number for a self-funded company: how long you lend money to each new customer.
• Good unit economics doesn’t mean a profitable company. You still need enough customers to cover fixed costs.

Sources: Jeff Jordan, Anu Hariharan, Frank Chen and Preethi Kasireddy, “16 Startup Metrics”, Andreessen Horowitz, August 2015 (GMV vs revenue, paid vs blended CAC, LTV based on profit, net churn); David Skok, “SaaS Metrics 2.0: A Guide to Measuring and Improving What Matters”, forEntrepreneurs.com (LTV:CAC above 3, CAC payback under 12 months and its exceptions, the gross margin assumption). All companies and numbers in the worked examples are fictional.

Related: Default Alive or Default Dead?, for the fixed-cost side of the same question; Why Most Small Service Businesses Charge Too Little, on fixing the margin before anything else; and Your First 10 Customers Won’t Come From Marketing, on the cheapest acquisition channel there is.