when do you know it's time to shard your database

we're trying to scale up our user service, and our one PostgreSQL instance is starting to show its limits. writes are getting slower, and reads are mostly fine but sometimes spike. sharding seems like the next logical step, but man, that's a huge architectural shift and adds a lot of operational work. are there any other ideas we should look at before jumping to sharding, especially for write scaling. when did you folks decide to ditch vertical scaling or read replicas for sharding. what were the main reasons that made dealing with sharding's complexity worth it