Customer lifetime value is often presented as a precise figure justifying how much customer acquisition can cost. In practice, it is a model with assumptions about retention, frequency, margin and time. A useful version makes that uncertainty visible rather than hiding it.
Choose the right unit
Decide whether you are analysing a customer, account, household or subscription. Remove duplicate identities where permitted and record changes in definition. LTV per order does not answer the same question as LTV per customer relationship.
Start with cohorts
Group customers by acquisition period, channel, product or another meaningful characteristic. Track actual behaviour over time. The overall average can conceal the fact that newer groups have entirely different quality.
Use contribution rather than revenue alone
Subtract variable costs, returns, discounts, service and other expenses associated with the relationship. A large revenue LTV can be misleading if growth is being bought with low or negative margins.
Show three scenarios
Calculate conservative, base and optimistic scenarios with clear assumptions. Avoid distant forecasts when you only have a few months of history. The decision needs to hold up across the adverse range, not just the most convenient outcome.
Connect CAC with payback time
The LTV-to-CAC ratio does not show when capital is returned. Add the payback period and funding requirement. Two channels with the same eventual LTV can have very different effects on cash flow.
Update when the system changes
Prices, the product, returns policy and channel quality shift the model. Keep versioned assumptions and compare the forecast with what actually happened. LTV is a learning tool, not a permanent customer label.
This uncertainty-aware LTV model is an original business framework developed by DigitalNow.