whens it ok to denormalize data in a microservices architecturetho

ive been thinking a lot about data consistency across microservices lately. while normalization is usually the default for relational databases, im seeing scenarios where denormalizing some data might actually improve performance and reduce inter-service communication. for example, if a user service needs to display a users name alongside their orders, should the order service store the users name redundantly? what are your experiences or thoughts on this trade-off in a microservices world