A brand calculates value at the time of conversion based on propensity and churn signals. Which value-centric category does this fall under?

This is a predictive value. The brand is not recording revenue that has already happened, and it is not applying a single averaged stand-in to every conversion. It is scoring each individual conversion at the moment it occurs, using signals about that customer's propensity to buy again and their likelihood of churning, to estimate what they will be worth over time. That makes the value personalised and model-driven, which is what distinguishes predictive from proxy. A proxy value applies the same calculated figure to every instance of a conversion type — every quote request worth £100, say — whereas a predictive value might score one signup at £40 and another at £400 based on their expected future behaviour. Predictive values are powerful for value-based bidding because they let the AI chase customers with high lifetime worth rather than whoever converts most cheaply today.