when to ditch the single database for a read replica

we've been running our app off a single PostgreSQL instance for a while now, and its been mostly fine. But as our user base grows, we're starting to see some slowdowns, especially on read-heavy operations. Im wondering, at what point does it make sense to introduce a read replica? Is it purely based on query performance degradation, or are there other factors like write contention or availability concerns that should trigger this decision? Curious to hear what others have found useful.