LTV is not a single number.

Choose a definition you can use

Start with a fixed time horizon and a value measure. For example, compare contribution margin per customer over the first 90 or 180 days after acquisition. Revenue alone can hide returns, incentives, servicing costs and other variable costs that shape the actual value of the relationship.

Compare like with like

Group customers by acquisition period and, where useful, first product, channel or market. Then track the value each cohort produces at the same age. A newer cohort will naturally have less observed value than one that has had a year to buy again; comparing their raw totals creates a false story.

• How many customers make a second purchase or deposit?
• How often do they return within the same observation window?
• What is the value after discounts, rewards and other variable costs?

Find the lever behind the number

LTV changes because underlying behaviour changes. Look at repeat rate, frequency, average order or transaction value and margin separately. If repeat rate falls, a larger basket among loyal customers may disguise a growing first-to-second-purchase problem.

Treat forecasts as forecasts

A predicted LTV can help allocate attention, but it is still a model. Keep observed cohort value alongside the forecast and check whether the estimate stays calibrated as customer mix, incentives and product experience change.