Unit Economics
How to calculate Customer Lifetime Value — the simple formula, the cohort-based approach, the predictive ML models, and which method is right for which kind of business. The math, the assumptions, and the operating implications.
Lifetime Value is the gross profit a customer is expected to generate across their entire relationship with the business. Done right, it's the most important number for determining how much you can afford to spend acquiring a customer.
Done wrong, it's used as marketing-deck inflation — a fictional number that justifies aggressive CAC spending without the underlying retention to support it.
LTV = Average Order Value × Purchase Frequency × Customer Lifespan × Gross Margin %
This works approximately for transactional businesses (DTC, retail, restaurants). It breaks down for subscription products because customer lifespan is hard to measure until customers have been around long enough to churn.
For subscription products:
LTV = ARPU × Gross Margin % / Monthly Churn Rate
This produces an asymptotic LTV — the expected value assuming the current churn rate continues forever. For early-stage products, this often over-estimates because early cohorts churn faster as the product matures. For mature products, it can under-estimate if retention is improving.
For more accurate LTV: build cohort revenue tables. Track each signup cohort's cumulative revenue per user over time. Project forward using the observed retention curve.
Example: if your 12-month-old cohort has produced $145 per user cumulatively, and the retention curve suggests they'll produce another $80 across the next 12 months and another $40 the year after, LTV at 36 months is ~$265.
This is more honest than asymptotic formulas because it's grounded in actual observed behavior rather than a single point-in-time churn rate.
Machine learning models can predict per-user LTV based on early-session behavior. Used heavily in mobile games, DTC subscription, and some SaaS contexts. The model is trained on historical data to predict which behaviors at day 1, day 7, day 30 correlate with high or low long-term LTV.
pLTV enables real-time CAC bidding — bidding more aggressively for users whose predicted LTV is high. Powerful when implemented well; misleading when the model is poorly calibrated.
LTV by itself isn't the goal. LTV in relation to CAC is. The conventional thresholds:
Using revenue instead of gross-margin LTV. Inflates the number 30–60%.
Including expansion revenue in LTV when modeling new-customer acquisition. Expansion is a separate motion; don't use it to justify acquisition spend.
Treating LTV as a fixed number. LTV evolves as cohorts mature, retention improves (or degrades), and the product changes.
Comparing LTV:CAC across categories with different payback profiles. SaaS LTV:CAC and DTC LTV:CAC mean different things — payback period matters as much as the ratio.