What is LTV (lifetime value)?

Why LTV governs acquisition

Acquisition cost has no meaning on its own. £50 to acquire a customer is either excellent or ruinous depending entirely on what that customer is worth.

LTV supplies the ceiling. Combined with CAC it produces the ratio most businesses actually manage against — and the payback period, which is often the more binding constraint because it determines how fast you can grow without running out of cash.

How it is estimated

• Historical. What comparable past cohorts actually produced. Accurate and backward-looking.
• Predictive. Modelled from early behaviour, so a decision can be made before the lifetime has elapsed. Necessary and less reliable.
• Cohort-based. Grouped by acquisition period and source, which is the only form that supports channel decisions.
• Net, not gross. After cost of goods, refunds and support, or it is a revenue figure wearing a value label.

Where LTV estimates go wrong

Extrapolating from the best cohort. Early adopters are systematically more valuable than the users a scaled campaign brings. Planning acquisition against their LTV is the most common way growth becomes unprofitable at scale.

Blending sources. A single account-wide LTV hides the fact that one channel produces users worth triple another. Decisions get made on an average that describes no actual user.

Ignoring the horizon. A 36-month LTV is not spendable today. Payback period constrains growth independently of how large the eventual value is.

Not revisiting it. Product changes, pricing changes and market shifts all move LTV, and models built on old cohorts keep authorising spend against a number that is no longer true.

LTV and AI-acquired users

Too early for reliable benchmarks, and the early signal is interesting: users arriving from AI assistant recommendations tend to be further through their decision, having already stated constraints and compared options in conversation.

That should produce better-qualified acquisition, and where volume permits it is worth cohorting AI-sourced users separately rather than blending them into a paid average. Our ARPU benchmarks for AI chatbot apps cover what is currently observable.