glossary · metrics
LTV (lifetime value)
Cumulative gross profit a customer generates from first purchase until they churn.
LTV is the financial proof that you built something both viable (customers pay) and lovable (they stay or expand). Interviewers do not ask about LTV to test arithmetic. They ask to see whether you pick the right formula for the business model, catch the gross-profit trap, apply cohort thinking, and connect the number to a product decision. Reciting “ARPU times lifespan, good ratio is 3:1” fails at any company that takes metrics seriously: it uses revenue instead of gross profit and pretends all customers behave identically. Both assumptions are wrong in ways that compound.
The formula hierarchy
Choose the formula that matches the product type. Move up as the business matures and data allows.
Tier 1: Simple subscription estimate. LTV = ARPU / Monthly Churn Rate. Fast and directional, but uses revenue, not gross profit. Back-of-envelope only.
Tier 2: Gross-profit corrected (the default you should use). LTV = Gross Profit per Customer / Revenue Churn Rate. This is what interviewers at growth-stage and public SaaS companies expect. Dropbox in 2015 spent roughly 67% of revenue on service delivery costs; a revenue-based LTV would show a number nearly 3x higher than the actual economic value. Use gross profit, use revenue churn (not logo churn).
Tier 3: Cohort-based LTV (what investors actually ask for). Track cumulative gross profit per customer cohort by time period: Day 7, Month 3, Month 6, Month 12. Jonathan Hsu at Social Capital made this a product-market fit signal: cohort LTV curves that “smile” (revenue per cohort grows over time as customers expand) are one of the strongest empirical signals that you built something people genuinely want to keep paying for. A curve that flattens at Month 3 points to a retention problem at the end of onboarding, not a pricing problem.
Tier 4: Compute-adjusted LTV (required for AI products). LTV = Compute-Adjusted Gross Profit per Customer / Revenue Churn Rate. For AI products, COGS must include inference compute, AI infrastructure, and support. Two customers paying $200 per month can have LTV of $2,500 (heavy user) versus $8,000 (light user). Traditional LTV treats them identically. The Inference Cost Efficiency Ratio (IER: revenue per dollar of inference cost) is an emerging proxy that AI PMs track alongside LTV to watch margin trajectory.
A worked example
ARPU $200, gross margin 60%, monthly churn 2%, CAC $1,800.
- Gross profit per customer: $120/month
- LTV = $120 / 0.02 = $6,000 (not the $10,000 revenue-based calculation, which is overstated by 67%)
- LTV:CAC = 6,000 / 1,800 = 3.3:1: healthy; the right call is to invest more in acquisition, not less
LTV:CAC benchmarks
| Ratio | What it signals |
|---|---|
| Below 2:1 | Model is broken; acquiring customers faster than you recover their value |
| 2:1 to 3:1 | Marginal; watch gross margin erosion closely |
| 3:1 to 5:1 | Healthy; median B2B SaaS at scale was 3.8:1 in 2026 |
| Above 5:1 | Often underinvesting in growth; unused channel capacity |
CAC payback period: best-in-class recovers CAC in 12 to 18 months. Above 24 months is a red flag regardless of the ratio. One caveat for 2026: per-seat pricing declined from 21% to 15% of SaaS in one year as hybrid models (base fee plus usage) reached 41% of the market. Fixed-ARPU LTV calculations are increasingly unreliable when ARPU varies month to month.
Which formula for which product type
| Product type | Recommended approach |
|---|---|
| Traditional subscription SaaS | Gross-profit cohort LTV (Tier 2 or 3) |
| Marketplace | Cohort gross profit using take rate, not GMV |
| AI / usage-based pricing | Compute-adjusted cohort LTV (Tier 4) |
| Early-stage, limited data | Tier 1 for direction; move to Tier 2 at Series A |
AI-first SaaS gross margins run 55 to 70%, compared to 78 to 85% for traditional SaaS. At the same ARPU, an AI product’s LTV is materially lower. Track whether IER improves over time as model efficiency gains compound.
weak
"LTV is average revenue per user times how long they stay. A good LTV:CAC ratio is 3:1." This fails on three counts: it uses revenue not gross profit, which can make a money-losing customer look profitable; it treats all customers as one average, hiding the fact that heavy users may destroy margin while light users generate all the value; and it recites a benchmark without connecting it to a product decision or a specific business model. Interviewers hear this from nearly every candidate.
strong
"Which LTV formula I use depends on the business model. For a traditional subscription SaaS, I use gross-profit cohort LTV: cumulative gross profit per cohort over time, not the simple ARPU/churn shortcut. That shortcut uses revenue and assumes all customers behave identically, both of which are wrong. For an AI product with variable inference costs, I compute LTV against compute-adjusted gross profit, because two customers at the same price point can have radically different unit economics depending on usage intensity. Concretely: if ARPU is $200, gross margin is 60%, and monthly churn is 2%, LTV is $120 / 0.02 = $6,000, not the $10,000 you get using raw revenue. With CAC at $1,800, that is a 3.3:1 ratio and a 15-month payback: healthy, which means I should invest more in acquisition. If I see a cohort LTV curve that flattens at Month 3, that is a retention problem, not a pricing problem. I would diagnose what is breaking at the end of onboarding before touching the pricing page."
LTV connects to product decisions at every level: which customer segments to invest in, which acquisition channels to fund, which features to build to drive expansion revenue, and when to kill a channel that cannot clear the payback threshold. The number is only useful when you can trace it back to a specific bet.
For related concepts, see CAC, churn, and cohort analysis. For the AI unit economics context, see LLM unit economics and proving viability.