when should you actually use a distributed tracing system? like, sometimes it feels like overkill, right?

ive seen more talk about distributed tracing lately, stuff like Jaeger and Zipkin. we run a ton of microservices, and figuring out problems across them is a real pain sometimes. i have seen systems with tracing and it does make finding bottlenecks and errors way easier. but it also adds overhead. when did you all decide you needed a distributed tracing system? what made your team or company finally pull the trigger