A value-based lookalike is like writing the spend next to every name on your regulars list. It asks Meta to treat a customer who spends heavily every month differently from someone who bought once.
It extends the standard lookalike. When you build a seed custom audience from CRM data you can attach an arbitrary numeric value to each user, and Meta uses it to decide, quantitatively, which users are worth more to you. There are three steps: create a value-based custom audience with is_value_based=1; fill it with identifiers plus the LOOKALIKE_VALUE schema key (any non-negative integer or float); and once the seed holds at least 100 people, create the lookalike with type custom_ratio. Reading, updating, deleting and targeting then work as for a normal lookalike.
Each ad account must accept the value-based lookalike terms of service. The number is a score for how much you want to resemble a customer, not revenue as such. Decide whether purchase amount, subscription length or margin is the right number first.
When to use
Use it when customer value varies widely and your CRM holds it; review seeds whose values are heavily skewed first.