Unit Economics: CAC, LTV, and Contribution Margin
Concept
A business can grow revenue every quarter and still be getting worse, not better, if each new unit it sells (customer, booking, transaction) costs more to acquire and serve than it brings in. Unit economics is the discipline of measuring profitability at the level of one unit rather than the aggregate — so growth is tested against "does more of this make us healthier?" instead of assumed to be good by default.
Three numbers do most of the work:
- Customer Acquisition Cost (CAC) — total sales + marketing spend over a period, divided by new customers acquired in that period. It's a blended cost: paid ads, sales salaries, tooling, all of it, spread across everyone it produced.
- Lifetime Value (LTV) — the total margin (not revenue) a customer generates over their relationship with the business, driven by how much they pay, how long they stay, and how expensive they are to serve.
- Contribution margin — revenue per unit minus the variable costs that scale with that unit (payment processing fees, hosting/compute per transaction, support cost per ticket) — the money actually available to pay back CAC and then contribute to fixed costs and profit. LTV is built on contribution margin, not gross revenue: a customer who pays a lot but costs almost as much to serve has a much lower LTV than the sticker-price revenue suggests.
These compose into the two health checks that actually get used: the LTV:CAC ratio (a commonly cited healthy floor is 3:1 — earn at least $3 of margin for every $1 spent acquiring) and CAC payback period (how many months of contribution margin it takes to recoup the acquisition cost). Payback period matters independently of the ratio because it's a cash and capital-efficiency question, not just a profitability one: a business with great LTV:CAC but a 30-month payback period needs a lot of capital to fund growth, since it's fronting the acquisition cost long before it's recovered.
Tradeoffs
| Metric | What it tells you | What it misses on its own |
|---|---|---|
| CAC alone | Cost efficiency of acquisition channels | Says nothing about whether acquired customers are worth keeping |
| LTV alone | Long-run value of the customer relationship | Easy to inflate with optimistic retention assumptions; ignores how much it cost to get there |
| LTV:CAC ratio | Whether acquisition spend is fundamentally sound | Hides when the payback happens — a 5:1 ratio over 5 years can still be a cash-flow problem |
| CAC payback period | Capital efficiency — how fast spend is recovered | Says nothing about long-run value past the payback point |
| Contribution margin | The real per-unit economics after variable costs, independent of scale | Ignores fixed costs — a business can have healthy contribution margin per unit and still be unprofitable overall if fixed costs are too high for its volume |
No single number is sufficient; the tradeoff table above is really an argument for tracking the ratio and the payback period and the underlying contribution margin together, since each one hides a different failure mode the others would catch.
When to use / when not to
- Compute unit economics before scaling any acquisition channel or pricing change — the question "if we 10x this, does it get better or worse?" only has an answer if the underlying per-unit numbers are known, not assumed.
- Segment by cohort or acquisition channel rather than computing one blended number for the whole business: a paid-ads channel and an organic/referral channel can have wildly different CAC, and blending them hides which one to invest in further.
- CAC payback period matters most for capital-constrained or early-stage businesses (it directly determines how much runway a growth push consumes); LTV:CAC ratio matters most for judging long-run channel quality once there's enough retention history to estimate LTV credibly.
- Don't treat unit economics as a one-time calculation — a cost driver (a new AI-inference cost per request, a payment processor fee change, rising ad costs in a channel) can flip healthy unit economics unhealthy without any change in the product itself.
- Don't compute LTV from revenue when contribution margin is available — revenue-based LTV systematically overstates value for any business with meaningful variable serving costs (payment fees, compute, support).
Common pitfall
Estimating LTV from optimistic retention curves extrapolated from too little data — a customer base six months old doesn't have three years of churn data, so any LTV built on an assumed long retention tail is a guess wearing the clothes of a calculation. This tends to produce unit economics that look healthy on a spreadsheet and turn out not to be once real multi-year retention data arrives. The more defensible approach is to use observed retention/margin over whatever horizon of real data exists, and treat anything projected past that horizon as an explicitly-labeled assumption rather than folding it into the headline LTV number unqualified.
Engineering Lens
Unit economics is a business metric, but engineering decisions move both sides of it directly: a feature that adds a per-request AI inference call, a new third-party API charged per call, or a payment processing integration all move the variable-cost side of contribution margin — and moving it the wrong direction can flip healthy unit economics unhealthy without a single pricing or marketing decision changing. The concrete habit this justifies: before shipping a feature with a genuine per-unit cost (not a fixed engineering cost, an ongoing marginal one), estimate its effect on contribution margin per transaction, the same way a proposed schema change gets estimated for its effect on query latency. The two questions ("what does this cost per booking now?" and "what does this cost per booking after the change?") are the engineering-facing form of a unit economics review, and they're answerable from the same telemetry a cost-optimization pass would already need.